Modelling Sociolinguistic Variation Historically
Despite the pioneering model provided by Romaine (1982a), systematic historical research on language change using multivariate techniques still has a long way to go. The limitations that any such undertaking will necessarily be confronted with have emerged from the previous chapters, with the 'bad-data' problem as their common denominator. In this section we will therefore only concentrate on the issues tackled and solutions adopted by our project. The previous chapters have shown that, despite many limitations, it is possible to reconstruct a number of social variables and relate them to language changes in earlier English. It is also possible to examine the interplay of external variables in long-term language change using the VARBRUL program. This can be done with high-frequency linguistic variables; on our list of 14 changes, there are five that provide enough data not only for cross-tabulation but also for detailed VARBRUL analyses (see 9.4.).
As to external variables, our analysis will include four: real time, region, gender and register. Region and gender have both yielded quite coherent results across time in the series of 14 changes we studied, suggesting that they are highly relevant to a sociolinguistic description of supralocal processes in Late Middle and Early Modern English. Of the other speaker variables, age and social rank will not be included because there would not be enough material in each cell to carry out the analysis. The variable that we have not yet discussed in detail is register, and it will therefore be introduced next.
Our view of register variation - and the use of the term register - can be associated with the model proposed by M.A.K. Halliday, which defines registers as varieties according to use. Registers are analysed in terms of the major dimensions of situations that can be expected to have predictable linguistic consequences: the field (type of activity and topic), tenor (addressee and other participant relations) and mode (channel of communication) of discourse (Halliday 1978; Eggins & Slade 1997: 47-58). In our case, the mode is written and the field is letter-writing, with a range of topics covered in each individual letter. What we will base our register analysis on are differences in tenor, i.e. social relations between correspondents. In this respect our analysis is well suited to be interpreted in terms of Bell's (1984) audience-design model.6
The CEEC encoding system records five basic addressee relations: nuclear family member (FN), more distant family member (FO), family servant (FS), TC (close personal friend) and T (other, distant recipient; see Nevalainen & Raumolin-Brunberg 1996b). For the purposes of the present analysis we have polarized the scale into two categories: family and close friends (FN, FO, TC) as opposed to all other recipients (FS, T). While defining kinship relations does not pose any major problems, defining close personal friends in the past may not always be self-evident. Terms of address, for instance, can be deceptive. Although friend is used throughout the Early Modern English period by close personal friends, social superiors could also address their inferiors as 'friends', and so could a merchant's wife her husband in the superscript of her letter, meaning 'a near relation', as in (9.1).
(9.1) To my loving frende John Johnson, be this delyvered at Callais. (Sabine Johnson, 1551; Johnson, 1161)
However, we could usually rely on the editor of a letter collection to single out those correspondents who counted as close friends in terms of minimal social distance and great mutual affection, often based on a long-lasting personal relationship, multiple shared interests in life, mutual confiding, and leisure time spent together. Friendship was often shown in the terms of address found in letter salutations (Nevalainen & Raumolin-Brunberg 1995). Nicknames would have been out of the question with distant recipients but were used by close friends. Charles Lyttelton did so when he informed his friend Charles Hatton about the death of his wife (9.2.).
(9.2) Deerest Kytt,
There is nobody in the world I can soe justly complaine to of the losse of your deare friend, because you were best acquainted how well I loved her and how much she deserved it. (Charles Lyttelton, 1663; Hatton I, 31)
Basically the three external variables of region, gender and register can be thought of as independent of each other. Although dialect levelling may be seen in writing earlier than in speech, personal correspondence cannot be expected to be homogeneous when it comes to the linguistic expression of register variation. Writing about early American English, Montgomery (1996) makes the assumption that Americans were multi-style speakers from the beginning, and also revealed this in their letters. The same assumption can be made about our writers, women and men alike. Evidence to support this will be discussed in section 9.3.
This approach will enable us to adopt Bell's (1984) hypotheses on audience design and to test the relative weight of social (i.e. regional and gender) variation as opposed to register variation in the letter corpus in the course of time. We have formulated these hypotheses in terms of an implicational scale as follows (the notation x > y meaning: 'the range of variation of x is greater than that of y'):
social variation > register variation according to addressee (> register variation according to auditors > overhearers > eavesdroppers) > register variation according to topic
Our analysis will concentrate on the first implication: social variation is greater and hence more fundamental than stylistic variation according to addressee in personal letters. We find this issue particularly relevant in the light of previous research, which suggests that the distinction between private and official correspondence correlates significantly with processes of language change (see 9.3., below).
Owing to lack of data from the lower social strata, we do not have direct access to the degree of register variation commanded by ranks at the lower end of the social spectrum. With our material it is therefore not possible to test the register hypothesis put forward by Finegan and Biber (1994, 2001) that lower-status speakers systematically favour processes of linguistic simplification and economy features, whereas higher-status speakers promote features of elaboration.
As demonstrated in Chapter 7, it is not, however, historically true that all processes of linguistic simplification spread from below in social terms. The middle and topmost ranks also promoted 'economy features' such as the subject form you. Moreover, with most changes studied here the more fundamental question arises of what actually counts as a process of simplification. The third-person singular -s was found to spread from the lower ranks, but it hardly constitutes a more economical suffix than -th. The ultimate economy solution would have been no suffix at all. The zero never generalized, although it was attested sporadically across the social spectrum in Late Middle and Early Modern English.