Problems and procedures in the construction of diachronic corpora: a case study of the passé composé and passé simple in Classical French

La construction d'un corpus diachronique: une étude du passé composé et du passé simple dans le français classique

University Of Georgia, 200 Broad Street, Athens, GA 30602, USA

Contact : magdi@uga.edu

Abstract

This paper describes a pilot study on constructing corpora to analyze grammatical change. The specific topic discussed in this study is the loss of the passé simple in French. This paper addresses the difficulties in using written data to discuss questions of language change, specifically when the forms in question are considered to be distributed differently in writing than in speech. The proposal of this paper is that comparisons between observed changes in different textual categories can illuminate the character of the change itself. Although the current paper is primarily concerned with how to ensure representativeness in diachronic corpora, the larger project addresses the character of the period of variation between the passé composé and the passé simple in French within the grammaticalization paradigm. The corpora described in this paper will ultimately be used to mount a multivariate analysis of the shift in past tense usage in French in order to determine if the passé composé was significantly favored in specific syntactic contexts. Additionally, the corpora will be used to determine if it is possible to quantify contexts in which the passé composé and the passé simple appear to be in variation with each other, i.e. to analyze the degree of semantic overlap between the two tenses as well as to identify whether or not both could be used to denote a series of discrete past events. In order to address these questions, it is first necessary to discuss possible methods for using written data to model language change. Thus the current paper specifically addresses the issues of text type sensitivity and stability in the construction of diachronic corpora, tackling the problems of data sampling and analysis, especially when there are restrictions on the availability of resources.