The basic method will be to extract examples systematically from existing databases of present-day English and Swedish. The historical part of the project will be based largely on existing studies, supplemented by use of historical corpora when required. Extracting relevant examples of -s and the NPs containing them is not straightforward. In an untagged corpus, the most ingenious string searches will still require a great deal of manual intervention to remove spurious or irrelevant examples.
In English the use of the apostrophe, both currently and historically, can be no more than one clue. Even with a tagged corpus, tagging error rates are always such that false positives must be removed. More serious is the problem of relevant examples missed because incorrectly tagged, particularly likely when -s occurs in an unusual context (e.g. the runner up’s entry, tagged in BNC as the contracted verb is). For the breadth and depth of study required to achieve our research aims, electronic searches have to be supplemented by time-consuming manual methods, to be carried out by the RA. We need accurate quantitative data for comparative purposes and comprehensive collections of marginal examples for qualitative analysis.
Furthermore, there may be issues of detail which the corpus examples cannot resolve. We therefore plan a programme of informant testing to tease out some detail as to the constraints on current acceptability. For this aspect of the project, we will collaborate with Professor Lieven in the School of Psychological Sciences, who has experience with eliciting linguistic data.
A workshop in Manchester, to be held two thirds of the way through the project, will help both to refine our ideas and disseminate them, as well as to widen the theoretical impact of the project by bringing related areas of research together.