Such preconceptions would make the analyst in

Info iconThis preview shows page 1. Sign up to view the full content.

View Full Document Right Arrow Icon
This is the end of the preview. Sign up to access the rest of the document.

Unformatted text preview: in form must result in a change of meaning. Tognini-Bonelli (2001) introduces a distinction she deems important in research involving corpora. She talks about corpus-based and corpus-driven studies. In corpus-based investigations the corpus is used only as a source of data which is supposed to support the hypotheses based on a preconception originating from sources other than the corpus itself (e.g. introspection or linguistic tradition). The corpus data cannot falsify a hypothesis; it can only corroborate it. In corpus-driven research the approach is different: The theory has no independent existence from the evidence and the general methodological path is clear: observation leads to hypothesis leads to generalisation leads to unification in theoretical statement (Tognini-Bonelli 2001: 84-85). The “observation leading to hypothesis” methodology can raise certain doubts. To start an observation the analyst must first ask a question – put forward a hypothesis, which could be tested against the data. This hypothesis is inevitably informed by the analyst’s knowledge of language, of linguistic theory, and of other social sciences. Such ‘preconceptions’ would make the analyst in question a corpus-based researcher. I can see no reason, however, why having these ‘preconceptions’ the analyst should be precluded from making sound data evaluation and from revising the original hypothesis, which would qualify as corpusdriven research. A corpus-driven approach is represented by Stubbs (2002 [2001]), who stresses the empirical nature of his work. He distinguishes between three types of linguistic examples: attested (retrieved from the corpus), modified (retrieved from the corpus but presented after a linguist’s intervention, e.g. shortening) and invented (based on the linguist’s intuition). Having made this distinction, in building his analyses and generalizing the results, he depends solely on the corpus data. Among other things, he analyzes the distribution and meaning of such lexical items as, for exam4 Lakoff – Johnson (1980: 126-138) also comment on the form-meaning relationship, when they refer to Bolinger’s (1977) conviction that a perfect paraphrase is not possible. They explain it by referring to spatial metaphors structuring our understanding of linguistic expression. Corpus linguistics and the language of mass media 57 ple, seek; large, big, great – series; cause vs. provide; undergo. In the study Stubbs identifies a number of lucid rules for conducting corpusbased lexical semantic research. He suggests that in order to avoid artefacts, the results received from one corpus should be checked against data from another. The frequency counts can be regarded as guidance to the interpretation of the meaning of words, but the infrequent collocations should not be altogether neglected, as an exclusive concentration on only the most frequent collocations may hide variation in the language. Second, collocations may differ quite sharply in different text-types. Many text-types are speciali...
View Full Document

This essay was uploaded on 02/24/2014 for the course LING 1100 taught by Professor Friedman during the Fall '09 term at Cornell University (Engineering School).

Ask a homework question - tutors are online