The importance of good networking: new methods for comparing sounds across individuals, accents and languages
Family trees have a long history in linguistics, and still offer a convenient and familiar way of indicating a particular kind of relationship between languages that are descended from a single common ancestor. However, there are well-known problems for the family tree model, involving similarities which do not reflect common ancestry, but instead arise from contact and borrowing, or from parallel change. In addition, trees usually stop at the level of individual languages like English, French, German or Welsh – whereas a lot of the exciting developments happen at the level of dialects, accents, or even interactions between individual speakers.
How, then, are linguists to talk about, model, and demonstrate the continuum between language variation and change? I shall argue that we increasingly need methods of comparison, models, and representations which are not specific to historical linguistics, dialectology, or variationist sociolinguistics, but allow us to work across all of these and consider common insights.
I will focus on a more neutral and flexible means of representation, involving networks. Network analysis is part of what we might call the ‘quantitative turn’ in studies of language variation and change. Quantitative approaches first establish a way of measuring or counting language features; then pass the results through a program such as NeighborNet. The output is a network which represents similarity, regardless of its source or cause. In this talk, I shall use results from the recent Sound Comparisons project (see http://www.soundcomparisons.com) to illustrate the use of quantitative methods and network representations for the whole continuum of comparisons – between languages; between accents; between individual speakers; and between different historical stages of the same language. This offers a framework for the reintegration of work on language variation and on historical change.
These methodological issues lead us to some specific questions which are of more general interest. Which varieties of English sound most alike? Which modern Englishes might be closest to the language of Shakespeare’s time? And are accents of English today getting more similar, or more different, or neither? However, these examples are necessarily limited, being based only on segmental phonetics, and on a single means of comparison and a single mode of displaying the results. We shall end with a more general overview of future needs and innovations for the field, including the development of new methods of quantification and representation across different levels of the grammar; an exploration of approaches to collecting data; and improved ways of sharing such comparative data.
Raum: KO2 F-180