The mental corpus: How language is represented in the mind. By John R. Taylor. Oxford: Oxford University Press, 2012. Pp. viii, 321. ISBN 0199290806. $99 (Hb). Reviewed by Natalia Levshina, Université catholique de Louvain

The book represents extensive argumentation in favor of the radically exemplar-based view of language, as opposed to the generative rule-based model. The core idea of the book is expressed in the metaphor of the mental corpus as a collection of our previous encounters with language, which constitute our linguistic competency. Although this idea is not new in usage-based linguistics, the book is probably a first systematic attempt to develop a fully exemplar-based theory of language representation. For empirical support, John R. Taylor uses a multitude of examples from contemporary English corpora and the Web. The book is written in clear, accessible language and can be recommended to students of linguistics and professionals alike.

The composition of the book is straightforward and serves the purpose of the main argument. In the beginning, the author presents the generative, or ‘dictionary plus grammar book’ (19), model which posits two essential kinds of linguistic knowledge: the list of words, on the one hand, and the list of rules to combine them, on the other hand. Chapter by chapter, T builds up a body of linguistic evidence that erodes this neat either-or distinction and charts the vast territory of exceptions and irregularities, which has been ignored by the generative theory. Starting with the units most resembling dictionary items (words and idioms), T moves on to more abstract schemas to reveal, on the one hand, a staggering amount of creativity in the lexicon and, on the other hand, an equally impressive portion of idiomaticity in what has been considered to be autonomous syntax.

By the end of the book, the reader should become convinced that ‘a very great deal, perhaps even the totality, of what occurs in language can be rightly said to be “idiomatic”’ (282). However, it is important to emphasize that T’s conception of language has nothing to do with simplistic behaviorism, since he shows that speakers are also capable of generalizations, which emerge from the instances and can be seen as probabilistic rules. This possibility of generalizations over data provides the basis for linguistic creativity and incremental innovation. A quintessential role in this process belongs to the frequency profiles of constructions encountered by speakers, who rely on this implicit knowledge in language processing and production.

The book demonstrates how the exemplar-based view of language representation and learning can provide insights in many research areas and topics, including lexical semantics, variational and diachronic linguistics, construction and cognitive grammar, language acquisition, and markedness theory in typology. Thanks to its timely and clear theoretical message and innovative contribution to various research domains, this book can be regarded as an important step in the development of the usage-based framework and linguistics in general.