Reviewed by Siaw-Fong Chung, National Chengchi University

This book collects six papers written as a tribute to John Sinclair to document some of his major concepts in lexicography and corpus linguistics. After an introduction to Sinclair’s career, Rosamund Moon discusses the Collins Cobuild English Language Dictionary (CCELD), which followed Sinclair’s principles of lexicography, featuring listing of senses (according to frequency), phraseology, and emphasis on corpora examples. M also details the conflict between lexicography and marketability, which is why the creativity of the Collins Cobuild project ‘did not lead to the kind of financial success which could have secured its survival in its original research-led form’ (2).

Geoff Barnbrook elaborates the two models of interpretation in Ch. 8 of Sinclair’s Corpus, Concordance, Collocations (Oxford: Oxford University Press, 1991): the open-choice and idiom principles. The first is a ‘slot-and-filler’ model that allows grammatical judgment to be made; and the second ‘presupposes that language users select from a set of semi-constructed phrases’ (23). Barnbrook then reinvestigates the distributions of the word back in a sample of 100 concordance lines from the Bank of English and concludes that the ordering of senses provided by Sinclair’s team is still predictive.

Charles Owen responds to ‘The Meeting of lexis and grammar’ in the same book, focusing on his analysis of of. Despite Sinclair’s claim to fuse lexis with grammar, Owen postulates a gradience analysis of the headedness of of (from a lot of money to a hatred of money to a child of money) and proposes an omissibility test for whether N1 or N2 is the head in each nominal group.

Using hatred as an example, Wolfgang Teubert argues that pattern grammar, which expects grammatical categories to be aligned closely with semantic distinctions, cannot account for the different meanings entailed by such phrases as hatred of a million coolies—whether coolies are the people who hate or those being hated. Instead, a ‘Continental model of valency and dependency’ (67) is proposed, involving a substitution test and dependency grammar. Using the traditional classification of genitives in Indo-European languages (e.g. genitivus partitivus, genitivus possessivus, genetivus explicativus), the author differentiates complements (their hatred for release) from adjuncts (their hatred for many years) (63).

Susan Hunston calls for attention to judgments of positive and negative ‘semantic prosody’, which have often been based on observations of positive or negative words surrounding the word of interest. As a counterexample, the word cause was found to express ‘no particular attitude’ when investigated in a specialized corpus of academic texts (87). Hunston warns that ‘collocational inference’, ‘intertextuality’ and ‘predictive value of frequent occurrences’ should not be exploited to the extent that one transfers attitudinal meaning from instance to instance (95–96).

The final article by Michael Toolan is entirely different, stimulating a philosophical, in-depth discussion of placing one’s trust in a text to revisit some of Sinclair’s assumptions. The author analogizes investigation of cooccurrences to placement of trust in a text. When one encounters an unexpected cooccurrence, ‘trust breaks down, does not truly exist, where the person or thing trusted turns out not to be trustworthy’ (108). Toolan also claims that no literary text resembles another since each writer possesses a unique writing style; therefore, sampling is impossible for literary texts. Finally, Toolan analogizes retrospection with cohesion, repetition, and reference.