Analysing sociolinguistic variation (review)

Variationist sociolinguistics is a highly method-driven discipline, in which apparently minor decisions about how to identify, count, or exclude data can drastically affect what are often presented as value-free empirical results. As such, it depends on its practitioners sharing techniques and assumptions about data treatment. Traditionally, however, these techniques have not been easily accessible. As Tagliamonte puts it, “it has often been noted that the practical details of how to actually do variation analysis are arcane, largely unwritten and, for the most part, undocumented” (p. ix). This book sets out to address this problem, as “one user’s tried-and-true manual of best practice” (p. x).

After a brief introduction (Chapter 1), which situates variationist sociolinguistics with respect to other branches of language study and lays out some of the major theoretical assumptions underlying the work, the book is structured as a step-by-step procedural manual. Chapter 2 (“Data collection”) deals with building a population sample that can be taken to be sociolinguistically representative of the speech community of interest. The focus here, as throughout the book, is on the need to let research questions guide the makeup of the sample. Chapter 3 discusses how the sociolinguistic interview can be used to generate relatively vernacular data, and how the judicious choice of questions and appropriate interview structure can create interactions approximating natural speech. Chapter 4 (“Data, data and more data”) offers tips on corpus constitution: keeping track of material, transcription, verification, and preparing corpora for automated searches and extraction.

Subsequent chapters discuss the analysis of variation within an existing corpus. Chapter 5 (“The linguistic variable”) explores the limits on what can be considered variable enough to analyze, and advises on the choice of a manageable variable to study. Chapter 6 addresses the issue of extracting tokens of the chosen variable, and “operationalizing” hypotheses from the literature as testable — code-able, measurable — claims. Chapter 7 starts with a brief defense of the concept of probabilistic modelling of language, then moves quickly into a description of the process of analyzing variation through multiple regression analysis using the Goldvarb program. This section is the core of the book, with approximately 100 pages spanning three and a half chapters. Tagliamonte sorts out the hundreds of day-to-day decisions that must be made during the analysis process to deal with problems of interaction, statistical significance, interpretation of constraint hierarchies, differences between distributional and multivariate analyses, and the like. Here, the reader is immersed in the behind-the-scenes cross-checking and data reconfiguration that are the basics of variationist work.

The book concludes with two useful chapters on how to turn findings (specifically Goldvarb outputs) into something that tackles the relationship between language and society. “Interpreting your results” (Chapter 11) evaluates methods of determining the strength of findings and using them to speak to sociolinguistic issues. “Finding the story” (Chapter 12) helps researchers turn their work into a presentable and comprehensible paper. The book includes exercises, which generally involve the application of each chapter’s advice to the reader’s own research, and a brief but useful glossary of terms.

Analysing Sociolinguistic Variation is a distillation of thirty-plus years of work in the variationist framework developed by William Labov and associated, in Canada at least, largely with work done in Montreal and Ottawa (where Tagliamonte first honed her craft). It assumes a very specific sociolinguistics model, which is both the book’s strength and its limitation. Those wanting to study vowel plots, gradience, or community of practice models will want to look elsewhere. On the other hand, for those wanting to find a community, conduct sociolinguistic interviews, transcribe data, choose a non-gradient linguistic variable, extract, code, and run it all through Goldvarb, this book will be a huge help. It is full of practical advice, procedural tips, and the academic equivalent of folk wisdom. Statements like “every sample design must be a balance between answering the research questions...