Previous Empirical Studies
As shown in Chapter 3, most of the historical research carried out within the variationist paradigm has been focused on genre and genre variation. The major generalizations to be made on the basis of this research are that external factors such as genre differences can indeed be reconstructed for the past and that they correlate with processes of linguistic change.
A number of facts have been systematically documented by using the varia- tionist approach. It has been shown, for example, by Raumolin-Brunberg (1991) that individual writers did vary their style according to genre. Sir Thomas
More's Noun-Phrase characteristics were shown to be sensitive to the sender/addressee relation, with a polarization between private (family) and official letters (most distant recipients).
A basic distinction between private and official correspondence has commonly been made, and contrasted with other genres, in studies based on the Helsinki Corpus of English Texts. The results typically indicate that linguistic innovations progress faster in private correspondence than in official. Kyto (1993: 124) shows the diffusion of the third-person -s to be the most advanced in private letters in the period 1560-1640, official letters and trial proceedings coming second, while diaries, histories and sermons lag behind them considerably. Kilpio (1997) observes that there are also features that connect official letters with written genres such as statutes. They are the two genres in which participial adjectives of the type 'aforesaid' were used most in Early Modern English (albeit their use was much more pronounced in statutory writings than in official letters).
The writer/addressee relations encoded in the CEEC allow more finegrained comparisons of social and register variables. Palander-Collin (1999) studied the use of the modal expression I think in the seventeenth century with respect to both gender and register with data from 77 women and 136 men. Her results showed that women used I think systematically more than men in all the recipient categories (FN, FO, TC, T). The male range - with close friends receiving the highest frequency - proved to be much narrower. In general, more variation was found in women's letters, with the highest frequencies occurring in the nuclear family category.7
Few multivariate studies have been carried out with historical data. Romaine (1982a) was the first to experiment with this methodology in her study of Middle Scots relativizers. Another early VARBRUL study is the analysis by Bailey et al. (1989) of the third-person singular present-tense suffixes in the fifteenth-century Cely letters. No external factors were considered in the study, which concentrated on the category and number of the subject.
Two other VARBRUL studies, Nevalainen (1991) and Schneider (1992), include both linguistic and external factors, but the former clearly dominate the analysis. Nevalainen (1991: 234-248) compared oral and literate genres promoting the exclusive adverbs only and but (Mary is only a child vs. Mary is but a child) in Early and Late Modern English. The analysis revealed a reweighting of this variable with both adverbs between 1500 and 1900. Schneider (1992) used the VARBRUL program to analyse the choice between who and whom in Shakespeare and also considered the external factor of text type (prose/verse). In his case study of whose vs. of which in Shakespeare and his contemporaries, Schneider included the range of authors as the external variable. In both cases, the external variable was found to pattern differently with the variant forms.
Both linguistic and external variables were also considered by Kyto (1997), who computed logistic regression analyses for the choice between be and have with past participles (he is/has come) covering the timespan from the fourteenth to the twentieth century. Apart from the author's sex and the formality of setting, the external factors selected were all textual: text type, relationship to spoken language and orality. Owing to the way the data were distributed, no more than two factors could be combined at a time, with the subperiod as one of the two in each combination. Text type and chronology were found to be among the more powerful external factors affecting the choice between be and have.8