International Journal of Corpus Linguistics - Volume 12, Issue 2, 2007

This paper discusses John Sinclair’s work in the field of lexicography by focussing on the first edition of the Collins Cobuild English Language Dictionary (1987), which was written within the Department of English at the University of Birmingham, and of which Sinclair was Editor in Chief. It provides theoretical and lexicographical background to the Cobuild Project, and reviews aspects of the first dictionary which were especially innovative, including its corpus basis, treatment of phraseology, and approach to the representation of meaning. It concludes by reflecting on the overall impact of Cobuild and Sinclair’s ideas.

This paper reassesses the description of collocation given by Sinclair in Chapter 8 of Corpus, Concordance, Collocation, in particular the implications of this description for models of language production and interpretation, contrasting the open choice and idiom principles. The concepts of independent and dependent meanings are explored, along with the relationship between texts and grammar. The arguments put forward in the Chapter are evaluated against collocation evidence obtained from a large reference corpus.

Sinclair’s grammatical work is notable for its strict reliance on performance data and its avoidance of psychological theorising. His argument rests on the observation that large corpora reveal a huge discrepancy between the predictions made by cognitive models of grammar and what actually happens in performance. This discrepancy cannot be explained away by appeal to encoding/processing deficiencies, but must be taken as reason to revise our view of grammar, especially by exploring its interdependence with lexis. In this he follows Halliday’s (1966) idea that every word has its own grammar. Sinclair’s 1991 paper on the word of (‘The meeting of lexis and grammar’) exemplifies his thinking in this area with particular energy and originality. Questioning the traditional classification of of as a preposition, he proposes a new, more semantically based, approach to the analysis of noun phrases traditionally said to contain postmodification of the head with an of-phrase, e.g. the horns of the bull. This chapter reviews Sinclair’s development as a grammarian, examines the arguments of the 1991 paper and suggests that while his work has been radically transformative in linguistics, it should not indefinitely avoid engaging with the problem of how best to represent what it is that we know when we say we know a language.

The view of pattern grammar is that syntactic structures and lexical items are co-selected and that grammatical categories begin to align very closely with semantic distinctions. While this is certainly a valid position when analysing the phenomenon of collocation, it does not really solve the problem for open choice issues. Not all language use can be subsumed under the idiom principle. The noun hatred, for instance, can co-occur with any discourse object for which hatred can be expressed. It can also co-occur with other lexical items standing for various circumstantial aspects. The grammatical structure itself often does not tell us whether we find expressed the object of hatred or some circumstantial aspect, as these structures tend to have more than one reading. Lexicogrammar, or local grammar, is more than equating a syntactic structure with a semantic pattern. We have to be aware of the different functions or readings a given grammatical structure can have. The framework of valency/dependency grammar can help us to make the necessary distinctions.

This paper considers the contentious term ‘semantic prosody’ and discusses a number of aspects of the concept described by the term. It is pointed out that although many writers use it to refer to the implied attitudinal meaning of a word, Sinclair uses the term to refer to the discourse function of a unit of meaning. Problems of apparent counter-examples, when a word or unit does not have the semantic prosody that is typical of it, are discussed. The second point made is that the phenomena described as ‘semantic prosody’ can be regarded as observational data, but that they are often used to explain subjective reactions to a given text or to predict such reactions. The issues raised by these different uses are discussed. Finally, the pitfalls of using concordance lines to observe attitudinal language in highly opinionated texts are discussed.

In this essay I celebrate and interrogate John Sinclair’s seminal paper, ‘Trust the text’, a paper in which several radically new ideas about the role of prospection and encapsulation in the reader’s processing of text are outlined. I mention some of the ways in which trust is fundamental to matters of language and cooperative communication, but also try to enlarge on what I think Sinclair has in mind. In reading on (and not re-reading), as we nearly always do when confronted with text, we are trusting the text in a more particular way, trusting it to have been composed in such a way that what follows will answer or complete what has gone before. This text-trust is perhaps the most fundamental structuring principle in written discourse, and mostly we apply it unwittingly; Sinclair’s paper broaches some lines of enquiry by which linguists might develop a fuller explanation of it.