In Talk that counts: Age, gender and social class differences in discourse, Ron Macaulay tackles two contentious but very different issues within Labovian Sociolinguistics. The first is the analysis of discourse level phenomena, specifically how to deal with ‘higher level’ variation within a quantitative paradigm. The second is Bernstein's (1971) restricted vs. elaborated code and the claim that middle class speakers have access to a more complex range of discourse structures when compared to lower class speakers. His findings on both are revealing.

In Chapter 1, Macaulay sets out the different approaches to discourse analysis, particularly with reference to the ‘functional’ vs. ‘formalist’ methods. He situates his methodology within the latter, where he is more concerned with ‘structure in focus’ as opposed to ‘dynamics in focus’ (Linell 2001:121). He states that the interpretation of use comes not from the analyst's ‘bias or misinterpretation’ (p. 11) but instead a ‘rather ascetic view’ (p. 11) that is gained from frequency correlations with the classic sociolinguistic categories age, gender and class.

Chapters 2, 3, 4 and 6 deal with data and method. Chapter 2, ‘Methodology’, classifies the different types of discourse features to be studied into three main types: (1) unambiguous forms such as the adverb very; (2) ambiguous forms which may have two meanings, such as you know; (3) complex forms such as passives or quoted dialogue. Macaulay points out that with an increasing level of complexity comes increasing levels of analyst intervention in the extraction phase with regard to what to actually count. After extraction, the ‘frequencies are expressed as the number of occurrences per 1000 words’ (p. 14). Chapter 3 details ‘The Sample’. The data come from Ayr in south west Scotland and Glasgow. Both are stratified by class (lower and middle) and gender, and Glasgow also has different age groups. It is noted that the data sets were collected in different ways: in Ayr, Macaulay conducted the interviews himself but in Glasgow the speakers conversed in peer pairs with no observer present. Chapter 6, ‘Talk in Action’, addresses this point, where Macaulay uses quantitative analysis to show that although the data were collected in different situations, they both result in ‘spontaneous conversation’ suitable for quantitative analysis. Chapter 3 also provides some basic frequency data and their correlation with age, class and gender. Total number of words per interview show that women talk more than men … but readers might like to note that this was ‘not quite statistically significant’ (p. 26). The complexities of Social Class as used as a measure in sociolinguistic research is the subject of Chapter 4. Macaulay concludes that as ‘it is unlikely that we shall ever employ methods and techniques that will satisfy the standard of sociological research … sociolinguists should not worry too much’ about this categorisation.

In ‘Decoding Bernstein’ (Chapter 5) Macaulay moves onto his second concern: Bernstein's (1971) work on discourse analysis and in particular his findings on restricted vs. elaborated code. He points out that despite the study having some major flaws, ‘it provides a useful starting point for the quantitative analysis of discourse’ (p. 43), as many of the claims can be tested empirically. This is exactly what Macaulay does in Chapters 7–12 in the analysis of discourse and social class.

Chapter 7 covers ‘Some Common Discourse Features' including oh, well, and in some detail you know, I mean and like. With you know and I mean, Macaulay finds a great deal of inter-speaker variability in use. However, the ‘differences in the functional use of these discourse features are probably more important than the overall frequency’(p. 86). For example, the class distinction reported by Bernstein – higher use of these discourse markers by lower class speakers – is not borne out in these data. Instead, there are differences in how some markers are used: the middle class speakers are more likely to use you know in a focusing function, whereas working class speakers use it ‘more as a bracketing feature’ (p. 86). Chapter 8 then turns to another of Bernsteins's claims, this time on ‘Syntactic Variation’: working class speakers’ syntax is characterized by ‘short, grammatically simple, often unfinished sentences, a poor syntactical construction with a verbal form stressing the active mood’ (Bernstein 1971:42). Macaulay focuses on a number of syntactic forms including coordinate clauses and ‘dislocated’ structures and finds that while there are age and gender differences, there are few, if any, significant differences in use between the two classes. In Chapter 9, ‘Moods and Modality’, Bernstein's claims regarding ‘complex verbal groups’ comes under scrutiny. However, Macaulay finds that ‘the lack of many social class, gender or age differences in the use of modal auxiliaries suggests that [they] are not a major contributor to differences in discourse style’ (p. 111). In contrast to the previous three chapters, Chapter 10 shows a significant correlation between ‘Adverbs and Social Class’: middle class speakers use a wider range of –ly adverbs such as briefly, clearly, supposedly when compared to lower class speakers. These results lead Macaulay to conclude that Bernstein's (1971:42) assertion that the lower classes show ‘rigid and limited use of adjectives and adverbs’ may have some support in these data. Chapter 11 on ‘Articles and Pronouns’ does not seem to test directly any of Bernstein's hypotheses but does find that there is a marked difference between males and females in the use of pronouns. Macaulay explains this with reference to topic: females talk more about people, males talk more about places. In Chapter 12 another gender distinction is found: females show higher rates of ‘The Use of Dialogue in Narrative’ ‘thereby dramatising the scene’ (p. 155). Macaulay also provides an analysis of the quotatives used in dialogue, confirming the rapid restructuring of forms ongoing in this area over the past two decades.

Chapter 13 summarises the ‘Results of Quantitative Measures’. Taken together, the results suggest that age is most likely to correlate with the discourse features under study, followed by gender then social class. For example, from a total of 42 measures of discourse features only 10 were statistically significant for social class. In the penultimate chapter, Macaulay discusses whether these results can be interpreted as a difference in ‘Discourse Styles’. The last chapter outlines 5 principles which make up ‘a heuristic guide for the investigation of discourse variation’ in future research (p. 190).

Talk That Counts makes two important contributions to the field of sociolinguistics. First, the study goes very much against the grain of current research in variationist sociolinguistics which tends to concentrate on the analysis of one linguistic variable and the detail of its patterning across social and linguistic factors. Instead, Macaulay tackles a whole range of discourse level phenomena which appeal to very different areas of the language faculty, encompassing discourse markers, higher level syntactic configuration and even narrative analysis. This approach provides an excellent overview of correlations in use across the classic sociolinguistic categories of age, gender and class, allowing for a more holistic view of similarities and differences in discourse use. Second, linguists have tended to be instinctively hostile to Bernstein's ideas of elaborated and restricted code. However, the claims have largely been combated with anecdotal counter-examples. Through quantitative empirical analysis, Macaulay demonstrates in black and white that the majority of Bernstein's claims have no basis in actual language use.

Macaulay points out that ‘the use of quantitative methods to investigate discourse variation is still at a very elementary stage’ (p. 190) and one of the thorniest problems which turns up again and again is circumscription of the variable context and the variants therein. It may be relatively easy to find, for example, the discourse marker you know in a stretch of discourse ‘but difficult, if not impossible, to find where they could have occurred but did not’ (Ito & Tagliamonte 2003:263). Or indeed, what else might be in its place. In line with many other studies, Macaulay employs the count per 1000 words approach and he shows that this method reveals statistical correlations of use with class, age and gender. But what might this methodology not reveal? For example, results show that the adolescents in the sample ‘show very little use’ (p. 85) of the discourse markers you know and I mean, leading Macaulay to hypothesise that ‘these speakers have not yet developed the full range of styles in the adult community’ (p. 85). At the same time, however, the analysis shows that the same age group use high rates of the discourse marker like. This might suggest not more limited range of styles in this age group, but simply that they are using another variant to do the work of more traditional ones. Thus, more detailed circumscription of the variable context may shed further light on the social and linguistic correlates of discourse level phenomena.

These methodological issues should not detract from an otherwise invaluable contribution to the field of variationist research. The breadth of coverage in this book provides an ideal baseline for future work on the quantitative analysis of discourse and is an excellent resource for anyone interested in talk that really does count.