The present collective volume consists of eight articles organized into three parts, viz. ‘The English gerund’, ‘Constructions and corpora’ and ‘Constructions and lexicalism’. The book is prefaced by a brief introduction by the editors entitled ‘Constructional approaches to language-particular description’. Moreover, additional introductions are provided for the three parts mentioned above: ‘Gerunds, categories and constructions’, ‘Corpus approaches to constructions’ and ‘Constructions and lexicalism’, respectively.

The introduction by Nikolas Gisborne and Graeme Trousdale focuses on two central claims of constructional approaches for language description, namely, the symbolic conception of grammar as consisting of conventionalized form–function correspondences, as well as the commitment to study language in its entirety, and not simply the core phenomena. Against the background of these shared claims, the editors pinpoint a number of relevant divergences in a fairly representative spectrum of constructional approaches. These include the degree of formalization, the stance on the (non-)compositionality of grammar, the functional pressures of language and the usage-based model. In turn, these issues are pivotal to the common themes that pervade the entire book, such as the connection between lexical items and larger constructions, the nature of gradience, the place of (complete/default) inheritance, the connection between synchrony and diachrony, the emergence of constructions, and a comparison of Construction Grammar (henceforth CxG) with other competing models, such as Head-driven Phrase Structure Grammar (henceforth HPSG), Optimality Theory (henceforth OT) and, most notably, Word Grammar (henceforth WG). This succinct but particularly clarifying introduction also spells out the motivation for the present collective volume: the recognition that constructional approaches are geared towards the investigation of important phenomena within individual languages rather than the establishment of a ‘universal grammar’ (p. 3).

The first part of the book opens with the paper by Bas Aarts, ‘Approaches to the English gerund’. Specifically, Aarts concentrates on English verbal gerunds of the type John having a sabbatical in sentences such as We were talking about John having a sabbatical from a synchronic standpoint. Rather than treating having as belonging to the class of verb and noun at the same time, Aarts invokes the notions of Intersective Gradience and convergence to provide a more robust account of otherwise problematic morphosyntactic properties of verbal gerunds in English. Aarts's alternative analysis, grounded on a thorough critical analysis of selected relevant literature, as well as an exhaustive analysis of data from the British component of the International Corpus of English (ICE-GB), differs from ‘mixed category constructions’ accounts in the following terms: ‘Fuzzy boundaries are to be avoided, and elements should not be allowed to belong to more than one category at the same time’ (p. 26). Aarts thus dispenses with the notion of a gerund (construction) and regards having in the above example as a nominal verb and teaching (as in We were talking about John's competent teaching of his students) as a verbal noun.

Graeme Trousdale's contribution, ‘Constructions in grammaticalization and lexicalization: Evidence from the history of a composite predicate construction in English’, rounds off the first part of this book. This paper addresses ‘the debate concerning the place of constructions in grammaticalization by examining the development of a particular subset of composite predicate (CP) constructions in the history of English’ (p. 33). Drawing on an exhaustive, bottom-up, synchronic and diachronic analysis of the syntactico-semantic behaviour of give-(nominal) gerund constructions (as in I gave him a kicking), Trousdale convincingly shows that these form part of a network of constructions (including the gerund and the ditransitive constructions). He then goes on to demonstrate how the constructional taxonomy developed by Traugott (Reference Traugott, Bergs and Diewald2008), involving macro-, meso- and micro-constructions at different levels of granularity, together with multiple default inheritance, can adequately capture the hierarchical relations among the constructions under analysis here at different levels of schematicity from a synchronic as well as a diachronic perspective. An important conclusion ensuing from this paper is that, given the emergent, non-fixed nature of constructions, diachronic CxG allows us to make descriptive and explanatory predictions about grammaticalization, as in the case of the highly schematic give-gerund construction, and lexicalization, as in more idiom-like constructions such as e.g. give rise to.

The second part of the book opens with a paper by Thomas Hoffmann, entitled ‘English relative clauses and Construction Grammar: A topic which preposition placement can shed light on?’. Drawing on converging evidence from the ICE-GB, as well as introspective grammaticality judgements subjected to statistical tests, Hoffmann investigates the constraints on the placement of prepositions in relative clauses, taking on board instances in which the preposition can precede the wh-relativizer (‘preposition pied piping’) (e.g. the place in which I live) or the relativized gap (‘preposition stranding’) (e.g. the place which I live in). Specifically, Hoffmann observes that stranding is more likely to occur with free relative clauses and (direct/indirect) questions, while pied-piping is more frequently attested with wh-relatives. At a higher level of granularity, the author demonstrates that the occurrence of stranding in free relatives and questions is not sensitive to text-type, while it is favoured in wh-relatives in informal texts. In concert with the usage-based approach endorsed in this paper, in which there is room not only for frequency, but also for processing and language acquisition factors, Hoffmann argues the case for a number of specific compositional constructions (or ‘micro-constructions’) to render descriptively adequate the constructional account of the phenomena under investigation here.

The contribution by Alexander Bergs (‘Shall and shan't in contemporary English – a case of functional condensation’) is a case study of the historical development of shall, shall not and shan't which provides empirical evidence for functional condensation (or ‘deconstruction’ in Bergs's terminology), that is, the gradual reduction of forms and functions of particular constructions and its interaction with ‘extralinguistic factors such as folk linguistic ideas, stylistic stigmatization and prescriptivism’ (p. 113). Bergs's exhaustive corpus analysis of the distribution of will and shall in several contemporary corpora of British and American English reveals that the difference between these forms is more prominent in American English (where shall is attested with first-person singular subjects). Moreover, the diachronic data allow the author to track the ‘deconstruction’ of shall, shan't and shall not and to conclude that while the first is eventually lost and replaced by should, the last two forms are still operational nowadays, albeit restricted to specific types of formal written discourse and to first-person interrogatives in spoken discourse. Bergs's overall conclusion is that a constructional approach making use of the Traugottian (2008) model can adequately account for the context-dependent, fine-grained semantico-pragmatic facets of the evolution of constructions in the history of language.

The second part closes with a paper by Julia Schlüter, entitled ‘Constraints on the attributive use of “predicative-only” adjectives: A reassessment’. The author analyses the role of semantic and phonological factors on the positional restrictions exhibited by ‘predicative-only’ adjectives (or, alternatively, a-adjectives), such as e.g. alive, asleep, aloof, etc. from a synchronic as well as diachronic standpoint. Drawing on a large-scale corpus, Schlüter provides a thorough classification and a quantitative analysis of disyllabic a-adjectives attested in pronominal position which underscores a recent gradual increase in attributive uses of allegedly ‘predicative-only’ adjectives. The semantic affinity of the predicative position with temporary, occasional properties in contrast to the permanent property typically encoded in the attributive position explains why a-adjectives are not generally fit to convey characteristic or identifying properties. As regards phonological factors, the sequence of an isolated a-adjective followed by a noun gives rise to an adjacency of two stressed syllables, commonly referred to as a stress clash. The rhythmic incompatibility of a-adjectives with the prenominal position explains why these adjectives are so rare in this position and why the presence of premodifying material as well as co-ordination, by allowing for a rhythmic alternation, may lead to rhythmically acceptable attributive structures. The author concludes that semantic and phonological constraints ‘have to be considered on an equal footing, with item-specific rankings rather than a predetermined priority, e.g. of semantics over phonology’ (p. 170).

The third part of the book begins with a paper entitled ‘Antitransitivity and constructionality’. Rosta's article focuses on the grammar of the English middle (e.g. This car steers easily) and anticausative (e.g. The window shattered) constructions. Rosta's initial claim is that these two constructions display a high degree of entrenchment (i.e. salience in memory of usage) and are fully compositional (i.e. semantically transparent and thus also predictable). He then goes on to make a distinction between ‘usage-constructions’ (‘u-constructions’ henceforth) and ‘grammar-constructions’ (‘g-constructions’). The former are defined by entrenchment, while the latter are characterized by non-compositionality, that is, some sort of idiosyncrasy about their meaning or form. The above two-fold distinction, as the author points out, is grounded on a divide between Grammar and Usage. As Rosta makes emphatically clear, this distinction has an ontological basis. In the author's own words:

Rosta then establishes a correspondence between the original definition of constructions revolving around the idiosyncrasy requirement (e.g. Goldberg 1995: 1) and ‘g-constructions’. By contrast, the recent, usage-based broadening of the original definition of a construction whereby ‘patterns are stored as constructions even if they are fully predictable as long as they occur with sufficient frequency’ (Goldberg 2006: 5, emphasis added to the original) fits in nicely with Rosta's ‘u-constructions’. In the remainder of the paper, Rosta provides a fine-grained analysis of the differences between middle and anticausative constructions. Thus, while middles feature an implicit agent, an adverbial and a habitual interpretation, anticausative constructions are not characterized by these properties. Rosta concludes that none of these semantic and pragmatic features are sufficient to consider them to be ‘g-constructions’; they should rather be regarded as ‘u-constructions’.

Nikolas Gisborne's paper, ‘Dependencies are constructions: A case study in predicative complementation’, is concerned with a comprehensive, unified account of raising (e.g. Jane expected Peter to be nice) and control (e.g. The landlord agreed to decrease the rent), small clauses (e.g. Jane considered Peter to be a fool/foolish) and non-verbal small clauses (e.g. Jane considered Peter a fool/foolish). As the title of the article suggests, one of Gisborne's main claims is that the dependency relations of WG are argument-linking constructions without phrase structure. He then illustrates this contention with specific reference to predicative complementation, which instantiates a construction type, with raising and control embracing the semantic areas of epistemic and deontic modality. The choice of predicative complementation as the object of inquiry is, as the author acknowledges, a challenge to the WG conception of dependency ‘as a simple pairwise relationship between two words . . . because it involves more than one dependency, and it involves an array of complex argument-linking information’ (p. 221). Gisborne further intimates that predicative complementation instantiates a network of dependencies (or, alternatively, a family of constructions) which can be adequately captured via the Traugottian inheritance model endorsed by Trousdale, among others, in this volume.

The third and final section of the book ends with a paper by Richard Hudson, ‘Word Grammar and Construction Grammar’. This contribution provides an illuminating discussion of the analogies and differences between WG and CxG, the main purpose of which is to propose four ideas from which CxG would considerably benefit. These are as follows: ‘that syntactic structure consists of dependencies between words’, ‘that semantic frames are part of the analysis’, ‘that the levels of phonology, morphology, syntax, and semantics are autonomous’, and ‘that the cognitive context should be enriched’ (p. 257). Hudson's point of departure is that these two models are strikingly similar, insofar as they are a bridge between cognitive and generative linguistics (pp. 258–9). WG and CxG are constructionist in a number of interesting ways: both models recognize that a person's linguistic knowledge ‘is a single unified network’ (p. 261, author's emphasis), that constructions are learned, that is, ‘constructed on the basis of the input together with general cognitive, pragmatic and processing constraints’ (Goldberg 2006: 3), that syntax is a different level of structure from semantics and phonology, and that linguistic study should be concerned with the entirety of each language, not simply with the ‘core’. However, an important difference concerns the ontology of each theory. Thus, while in CxG the full gamut of linguistic elements ranging from morphemes to (complex syntactic) patterns (e.g. the sooner, the better) are regarded as constructions, WG only posits ‘dependencies between individual words’ (p. 263, author's emphasis). Hudson provides a number of compelling arguments for dependency structure from different areas, such as computational linguistics and child language, as well as some advantages for linguistic description. A second important difference is that WG, unlike CxG, can furnish richly articulated and precise semantic analyses by incorporating the entire semantic frame into the semantic structure. Moreover, CxG assumes that there is one level of ‘form’ (i.e. syntax) in contrast with the multi-level analysis invoked in WG, in which linguistic structure is taken to involve a number of levels including phonology, morphology and syntax. Hudson then categorically denies the existence of constructions as the basic units of analysis as follows:

Finally, while praising the psycholinguistic orientation of the Goldbergian formulation of CxG (Goldberg 2006), Hudson pinpoints three elements which could help maximize the cognitive context of CxG: spreading activation (i.e. activation takes place in the individual node rather than in the construction), universal inheritance hierarchies (i.e. inheritance hierarchies apply to all kinds of entity) and best-fit binding (i.e. a combination of spreading activation with binding and default inheritance).

In what follows, I will provide a brief assessment of this collective work. In the blurb at the back of the book one can read the following:

In my opinion, the present volume admirably succeeds in engaging the reader in a truly illuminating and challenging reappraisal of CxG for the description of different areas of English grammar, including, as rightly pointed out by the editors (p. 173), levels of analysis often left unaccounted for by most construction grammarians, such as phonology (the articles by Hoffmann, Bergs and Schlüter). This collective work offers new insights into relevant long-standing phenomena, such as the behaviour of nominal and/or verbal gerunds (the articles by Aarts and Trousdale), as well as yet unexplored issues, such as the demise (or ‘deconstruction’) of shall (the article by Bergs) and the emergence of attributive uses of ‘predicative-only’ adjectives (cf. the article by Schlüter), among others. The contributions in this book indeed excel at being critical, showing how different concepts and formalisms impinging on substantial theoretical issues can be insightfully combined with the generalizations emerging from a thorough examination of naturally-occurring corpus data alone or in conjunction with psycholinguistic evidence. Relevant cases in point are the boundaries between grammar and use, the operational definition(s) and boundaries of ‘constructions’, and the type and role of inheritance taxonomies, among others. Thus, while sympathetic to a bottom-up, usage-based approach, Rosta and Hudson forcefully vindicate a divide between grammar and usage. Moreover, a default inheritance with overrides is fully subscribed to in most of the contributions in this book. However, some authors (e.g. Bergs) invoke the formalism of HPSG, while others advocate a combination of CxG and OT (e.g. Schlüter). The most challenging assumption is perhaps the distinction sanctioned between non-compositional constructions (‘g-constructions’) and compositional but (highly) frequent ones (‘u-constructions’). From a theoretical viewpoint, Rosta is agnostic on the need to postulate ‘g-constructions’ (p. 192), while Hudson takes a much more radical stand and rejects outright the central CxG tenet ‘that there is nothing in language except constructions’ (p. 286). From a descriptive perspective, the papers by Aarts and Trousdale illustrate two different positions: for the former, (nominal) gerunds are not considered to be constructions, while (verbal) gerunds are by contrast granted a constructional status. As also stated in the blurb, the present volume constitutes a formidable example of how synchrony and diachrony can be integrated in a constructional approach. In particular, the application of the Traugottian model (2008) that pervades to a greater or lesser extent most of the chapters in the book is certainly most welcome.

I should now like to single out some points which, to my mind, could perhaps have been elaborated a bit further, or which still await further negotiation with CxG. These should not, however, be taken to detract from my very positive evaluation of the structure and content of the entire book.

First, the assumption that a clear-cut dividing line should be sanctioned between grammar (or competence) and usage (or performance), although discussed at length by Rosta, is at odds with the standard definition of a usage-based model, as currently understood in functional, cognitivist and/or constructional approaches (see Gonzálvez-García and Butler 2006: 83). Therefore, a working definition of a usage-based model, as understood by the contributors to this volume, should have been spelled out.

Second, the rejection (Hudson) or questioning (Rosta) of the original formulation of constructions (Goldberg 1995: 1), understood as non-compositional form-function pairings, clashes with one of the foundational claims that gave the initial impetus to CxG, namely, that syntax is by and large motivated by semantics. In fact, one of the great achievements of CxG over the last twenty years has been precisely to demonstrate, pace Chomsky, that syntactic patterns carry meaning in themselves in relative independence of the lexical items occurring in them (cf. Goldberg 1995, 2006, inter alios). Moreover, Boas (2009) has recently argued that both definitions of constructions (as non-compositional as well as compositional but highly frequent correspondences) need to be retained in CxG, because these are complementary to fully understand the dynamics of communication: the former definition is needed for decoding, while the latter is necessary for encoding. However, to be fair to Hudson and Rosta, it should be noted that some models (e.g. Halliday's Systemic Functional Grammar) reject the notion of the construction as a strict pairing of form and function, while still adhering to the tenet that syntax is motivated by semantics (see further Butler 2006 on semantic motivation).

Third, in his constructional account of predicative complementation, Gisborne (p. 221) affirms that ‘[t]he model requires a view of the relationship between dependencies (or constructions) and lexical items which is essentially that of Croft (2003)’. In view of the fact that, while WG only recognizes dependency relations, Croft (2003) claims that constructions (understood either as non-compositional or fully transparent correspondences) exist, the connection between WG and Croft (2003) should have been worked out in some detail. As a matter of fact, Radical Construction Grammar (RCG) (Croft 2001), and in particular the claim that argument structure is not only language-specific, but also construction-specific (see also Goldberg 2006: 225–6), has important consequences for a constructional investigation of individual languages (e.g. English), which is the main theme of the book.

The present volume undoubtedly holds a great promise of not only making it possible, as Hudson has put it, ‘for ideas to flow relatively smoothly’ (p. 262) between WG and CxG, but hopefully also making it possible among adherents of other models in the present-day functional-cognitive scenario.