This book addresses a number of long standing and still largely unresolved issues in lexical semantics. These centre on the observation that the semantic value of a word varies according to its context of use. Open refers to different activities in open a book, open the office, and open the curtains; flying a plane is different from flying a kite, which is different yet again from flying a flag; in designates different relations in the kitten in the box, the flowers in the vase, and the light bulb in the socket. Some of this variation might be accommodated by supposing that a word may be polysemous. However, the phenomenon of meaning variability is so pervasive that it would be impossible in principle to list all of the possible meanings of a word in all of its possible contexts of use.

The following questions arise: (i) How is the meaning of a word to be stated, abstracting away from the specifics of its contextualized uses? (ii) How is a contextualized meaning to be derived? (iii) In case the word in question is regarded as polysemous, what are the criteria for distinguishing the different senses? and (iv) How is the variability of word meaning to be reconciled with the compositionality principle – the process whereby the meaning of a larger expression is built up from the meanings of its parts?

Evans tackles this cluster of issues by making a clean distinction between linguistic and non-linguistic, or conceptual knowledge. Meaning emerges through an interaction of these two independent systems and is a property, not of linguistic units stored in long-term memory, but of context-bound utterances. Words, as such, do not have meanings in the sense that they are associated with fixed chunks of conceptual content which they contribute to the compositional process. Rather, a word provides access to conceptual knowledge (the cognitive models of the book's title). Cognitive models comprise coherent systems of knowledge, derived from experience; they are multi-modal in nature and extensively cross-referenced, ranging over individuals, types, events, and properties. Understanding a linguistic expression involves the selective activation and fusion of facets of cognitive models (the meaning construction of the title) in accordance with the syntactic constituency of the expression. The result is a coherent conceptualization, or ‘conception’, as Evans prefers to call it.

Although a word lacks conceptual content, it is nevertheless associated with a bundle of stable and purely linguistic information (or several such bundles, in the case of polysemy). These are the lexical concepts of the title. There are three components to a lexical concept. The first, and the one that is given most attention in the book, pertains to highly schematic content derived from a word's grammatical category. The linguistic concept associated with red is simply ‘property of a thing’, in keeping with the status of the word as an adjective. The conceptual content ‘redness’ resides in the conceptual system and our specific memories of red things, such as fire engines, red ink, and red squirrels, any one of which might be activated by a specific use of the word.

A second facet of a lexical concept pertains to selectional preferences, referred to as a ‘lexical profile’. A lexical profile concerns the kinds of items a word typically occurs with and the kinds of syntactic environments it typically occurs in, in other words, its collocations and colligations. Lexical profiles are an important criterion for diagnosing polysemy, the situation where a word is deemed to be associated with more than one lexical concept. A third facet, which is only briefly discussed in the book, is referred to as ‘pragmatic point’. This has to do with the speaker-, register- and discourse-specific values of a word.

Evans's approach to syntax is broadly in line with construction grammar accounts. Accordingly, it is not only words that are associated with lexical concepts but also morphemes, multi-word expressions, constructional idioms, and syntactic configurations more generally. Thus, a transitive clause, a passive verb phrase, predicative clauses, and so on, are associated with schematic semantic content which is fleshed out by drawing on the cognitive models accessed by their constituents. The compositional process roughly follows an expression's constituency, with the innermost units being interpreted first.

The book consists of sixteen chapters, arranged in five parts. The four chapters of Part I (‘Introduction’) address the problems raised by the semantic variability of words, introduce the basic architecture of the theory, and offer a brief introduction to Cognitive Linguistics (of which the present theory is meant to be an instance). A central feature of Langacker's (Reference Langacker1987) theory – the claim that semantics is encyclopedic in scope – is however rejected. The six chapters of Part II (‘Lexical representation’) deal with lexical concepts and the conceptual models which they access. Chapter 8 provides a case study of ‘Polysemy’, examining ‘abstract’ uses of the prepositions in, on and at, while Chapter 9 gives an account of ‘Cognitive models’, drawing heavily on the grounded cognition theory of Barsalou (Reference Barsalou2008) . The three chapters of Part III (‘Semantic compositionality’) address the process of semantic composition. There are three stages: selection of lexical concepts (in the case of items which are polysemous); integration of lexical concepts into larger units; and conceptual interpretation, resulting in a context-bound conception.

Part IV (‘Figurative language and thought’) exemplifies the model on two case studies: non-literal expressions and temporal expressions. Chapter 14 proposes a rather traditional account of ‘Metaphor and metonymy’, which sees interpretation as the resolution of a conceptual clash. Chapter 15, ‘The semantics of time’, takes issue with the widely held view that time is, of necessity, conceptualized metaphorically in terms of space. It is argued that time can indeed be conceptualized independently of spatial notions. Conceptual metaphors, popularized by Lakoff & Johnson (Reference Lakoff and Johnson1980), are located in the conceptual system and allow conceptual knowledge pertaining to one domain to be recruited for the interpretation of another domain. The final chapter, ‘Conclusion’, which constitutes Part V of the book, promotes the theory of lexical concepts and cognitive models (LCCM) as a ‘joined-up’ synthesis of dominant trends in Cognitive Linguistics.

The book is an impressive achievement, not least because of the detail with which the theory and the various principles and conditions which are built into it have been worked out. The case for cognitive models, and their role in semantic interpretation, is particularly well made. However, I find an overall evaluation difficult to make. One reason stems from the extremely limited range of data which is used to ground the theory. Evans insists on meaning being a property of contextualized utterances, not of sentences, even less of words. It is remarkable, therefore, that the data on which the theory is illustrated are all – barring such things as a slogan observed on the back of a big red bus in Brighton (275) – invented examples, constructed to make a point. Even discussions of lexical profiles make no reference to corpus-derived data. What is more, we find the very same handful of example sentences (referred to throughout as ‘utterances’) repeated and recycled throughout the book.

The focus on a small number of invented sentences can lead to ad hoc and even false analyses. Given the role of lexical profiles in the analysis of polysemy (141f.), it would surely have been in the spirit of the theory to have let the various senses of a putatively polysemous word emerge on the back of usage data, as exemplified in, for example, Gries & Stefanowitsch (Reference Gries and Stefanowitsch2006) . Evans, on the contrary, bases his polysemy accounts on a priori intuitions about the conceptual distinctiveness of various uses (142). These are then tested on the distributional restrictions exhibited in invented examples. Thus, in the discussion of flying, a number of lexical concepts are hypothesized, one of which is [self-propelled aerodynamic motion]. (The practice of referring to lexical concepts by paraphrases in small capitals raises the question of the ontological status of these glosses, a topic which is not addressed in the book. In keeping with the theory, the conceptual content of the gloss cannot be part of the lexical concept.) The singling out of this concept is justified by the claim that only self-propelled entities can exhibit aerodynamic motion, as shown by their occurrence in the frame [__ is flying (in the sky)]. Birds and planes can ‘fly in the sky’ but not, we are told, tennis balls (147). This is simply wrong. Here are some examples of flying balls from the British National Corpus:

1. (1)

   1. (a) And as a television sport it [=golf]'s also hard because all you see is a ball flying through the air and you have no sense of scale or depth

   2. (b) Because a wet ball will fly farther, use less club then [sic] normal on approach shots

   3. (c) Table tennis is something else. The ball flies across the table.

The most strongly data-driven chapter is Chapter 8, on ‘state’ senses of the prepositions in, on and at. Evans's account would have been more compelling if it had been prefaced by an independently derived data set of state expressions against which the analyses could have been validated and, possibly, refined. According to the [psychosomatic state] sense of in, a person can be in love, in shock, in despair and in ecstasy (165). Why, then, can a person not be in euphoria (the appropriate usage is be in a state of euphoria)? The ‘messiness’ of actual usage is simply ignored. There is, for example, an only partial congruence between predicative and adverbial uses of the prepositional phrases. While you can nod ‘in approval’, you cannot ‘be in approval’; you can answer ‘in surprise’, but you cannot ‘be in surprise’. These restrictions are not discussed, or even acknowledged.

It might also be queried whether a listener, in interpreting the expression in love, really does select the lexical concept [psychosomatic state] out of the many which are associated with the preposition. Surely, (be) in love is a fixed expression, learned as such. It may instantiate a more general pattern, but it is not interpreted compositionally. The same might be said of other cases illustrating the ‘protean’ nature of word meanings. To be sure, open refers to different kinds of activity in the three expressions cited at the beginning of this review. Might it not, however, be an option to suppose that language users have simply learned these phraseologies and their associated conceptual content? On this view, the question of what open, abstracted from its uses, actually means, does not arise.

The paucity of data sometimes leads to what I think are unjustified generalizations. The discussion of metaphor in Chapter 14 is based on only two (!) examples, one of which (The student's grades went up), on Evans's own criteria (299), is probably not metaphorical at all. (‘Increase in value’ is a conventionalized sense of go up.) Be that as it may, both metaphor and metonymy are said to trigger ‘clash resolution’. With reference to the examples My boss is a pussycat and the old favorite The ham sandwich wants his bill, it is proposed that in the case of metonymy it is the ‘topic’ of the sentence which is understood figuratively, whereas in metaphor it is a non-topical element (293). I doubt whether the grandly named ‘Principle of Context-induced Clash Resolution’ (292) applies beyond a small set of selected examples. It certainly would not apply to the following quasi-minimal pair:

1. (2)

   1. (a) This organization needs a new face.

      (metaphor: ‘face’=‘image’)

   2. (b) This organization needs some new faces.

      (metonymy: ‘faces’=‘new employees’)

Many scholars have attempted to demarcate linguistic and conceptual knowledge, placing the boundary at different places, on different criteria, and for different purposes (Geeraerts Reference Geeraerts2010). Evans's approach is a radical one, seeking to eliminate virtually all conceptual knowledge from the linguistic system. Conceptual knowledge is merely ‘accessed’ by lexical concepts, it does not inhere in them. Exactly how the accessing process works is not really addressed. What seems clear, though, is that in order for lexical concepts to be able to link up with conceptual knowledge, cognitive domains need to be already ‘indexed’ by linguistic labels. How else would a speaker know which ‘property of a thing’ is to be accessed by the word red? Conversely, a speaker who wishes to state the colour of a thing needs to know which lexical concepts are available for this purpose. Evans seeks to evacuate conceptual knowledge from the language system; it is questionable whether language can be evacuated from the conceptual system.

Given the need for the indexing of words to conceptual knowledge, and vice versa, one might well ask what is to be gained from insisting on a strict separation. The core mechanics of Evans's theory, concerning conceptual integration and interpretation, are not, it seems to me, dependent on this distinction; in fact, they might well be compatible with a fully encyclopedic view of lexical semantics. We need not thereby endorse the clearly fallacious notion that words are associated with fixed invariant packages of conceptual information, nor, for that matter, the equally dubious proposition (71f.) that words are represented in the mental lexicon by strings of phonemes (Port Reference Port2007). In fact, exemplar theories of phonological representation (Pierrehumbert Reference Pierrehumbert, Gussenhoven and Warner2002) could provide a model for semantic representation, namely, in terms of a cluster of remembered instances of how a word has been used, and with what semantic values. The ‘illusion of semantic unity’ (207), and indeed of phonological unity, could result from the fact that a word, or larger expression, contemplated in isolation, may tend to evoke a default context of use.

The theory of lexical concepts and cognitive models strikes me as very much a work in progress. As the book nears its end, the prose becomes increasingly repetitive and turgid and some authorial editing would, I feel, have been appropriate. It remains to be seen how the theory will be able to tackle, in fresh and insightful ways, the complexity, not to mention the messiness of actual usage, and what it can contribute to the study of the mental representations which are abstracted from encountered usage, and which, in a feedback loop, sanction a speaker's own verbal behaviour.