Speech Rate, Pause and Sociolinguistic Variation

Studies in Corpus Sociophonetics

T. Kendall

• Publisher: Palgrave Macmillan
• Year: 2013
• Print ISBN: 9781349320950, 9780230249776
• eBook ISBN: 9781137291448
• Topic: Languages & Linguistics
• Subtopic: Phonetics & Phonology
• Index: Languages & Linguistics

This book provides a fascinating account of the psycholinguistic and social factors behind variation in speech timing in US English. With detailed discussions of its methods and data, it also acts as a valuable model for conducting corpus (socio)phonetic research.

Part I
Speech Rate, Pause, and Corpus
Sociophonetics

1

Looking Forward

1.1 Introduction

This book is about speech timing and, more specifically, about variation in the temporal features of speech rate and silent pause in spoken American English, as viewed from a quantitative sociolinguistic, and to a lesser degree psycholinguistic, perspective. Although it is a book explicitly about the sociolinguistics of speech rate and pause, it is also a book more broadly about corpus-based methodologies and about conducting large-scale sociophonetic research. Throughout this book, I attempt to give as complete an overview of the corpus-based methods and statistical maneuvers I employ as I can. As such, I also provide many resources connected to this book on its website – http://ncslaap.lib.ncsu.edu/speechrateandpause/ – including electronic versions of some data files and tools for, for example, counting syllables in English language orthographic transcripts. It is – of course! – my hope that this project contributes towards our substantive understanding of patterns of speech timing in human language, but I also hope that readers find it useful as a guide to doing large-scale, quantitative sociophonetic research.

In many ways, this book is also about recycling older sociolinguistic recordings and mining them for new phenomena and for the exploration of new questions. It follows from a thread of my research on corpora and data in sociolinguistics (Kendall 2007a, 2008a, 2009, 2011, forthcoming a, b). In particular, while I do not intend this book to be a revision of my PhD dissertation (Kendall 2009), it picks up from that work. There, I discussed in detail the Sociolinguistic Archive and Analysis Project (SLAAP; a web-based sociolinguistic data management system I built at North Carolina State University)¹ and meta-theoretical questions about data, their treatment, and representation in sociolinguistics, and then turned to a preliminary examination of speech rate and pause as an exploration of how the approach to data implemented in SLAAP made such investigations possible. After several years of continuing to think about and study variation in speech timing, this book presents a much more focused and complete treatment of the sociolinguistics of speech rate and pause. Other than a brief overview of the relevant background in Chapter 3, I leave the larger meta-theoretical discussions of data and data management to the other outlets listed above.

With the goal of examining speech timing in depth, as indicated by the book’s title, I limit my focus to patterns of SPEECH RATE and SILENT PAUSE in human language. Other temporal factors, such as segmental durations and speech rhythm, are of interest – and hopefully illuminated upon by the specific foci of this project – but for the sake of maximizing depth in my coverage, I do not pursue them in any explicit way. Pauses, both silent and filled (e.g. uh, um), are extremely interesting from a number of perspectives, but I will maintain a view on pause from a temporal perspective, focusing more on pause durations than on other potential areas of research, such as the clausal location of pauses, the frequencies of pauses, or the semantics of filled pauses.² (Although I will from time to time touch on these subjects, for example, by considering the role of pause location and pause type on silent pause duration in §6.5.)

While this book focuses closely on speech rate and pause, it is also a book more generally about where we find socially differentiated linguistic behavior, the STRUCTURED HETEROGENEITY of Weinreich, Labov, and Herzog (1968). It is about what variation in language can be accounted for by readily enumerable linguistic and social factors. It is about how much of the apparent messiness of variable temporal features – specifically the rate at which speech is uttered and the length of a mid- sentence pause – can be modeled thanks to the burgeoning quantitative and statistical techniques available to the social sciences of the early twenty-first century. At the same time, this book is about what cannot be modeled in this way. It is about what variation is unaccounted for in a large-scale corpus-based analysis, and, better yet, what light we can shed on the processes at work in language production from the unaccountable bits.

Importantly, it is a book about doing CORPUS SOCIOPHONETICS. In these pages, I ask what new things we can learn from treating the large collection of sociolinguistic recordings housed in the SLAAP archive, which were originally collected for various, unrelated sociolinguistic projects, as a coherent sociolinguistic corpus. And I ask the broad question of what the large amount of data obtained through corpus-based analysis (here, ~30,000 measurements of each of the dependent variables) gets us that a smaller dataset does not. Do we learn more from 1000 tokens of a variable from each speaker than we do from, say, five, or from a single estimation of each speaker’s general tendency?

Over the past half-century, sociolinguistic research has collected a huge amount of naturalistic speech data. Typically,³ these data have been used by their collectors to investigate specific research questions and then, after active use over the course of some period of time, the data are put aside and new data, from new communities and research sites and with an eye to new questions, are collected. In recent years, there has begun to be a change in the way that sociolinguistic data are collected and conceived across the discipline. Partly, this is a result of an increasing ability for and interest in conducting REAL-TIME research on language change – that is, to examine comparable data from multiple points in time to examine language change (see Bailey 2002, Sankoff and Blondeau 2007, Gregersen 2009). But, partly, this is a more general result of a reconsideration of sociolinguistic data as corpora (cf. Beal, Corrigan, and Moisl 2007a, b, Kendall 2008a, 2011). Along with the growing sense that sociolinguistic recordings are useful in the long term is a growing sense that they ought to be more “public” than in the past. As Gerard Van Herk and I wrote: “The previous, dominant model of considering sociolinguistic data as too valuable to ‘part with’ or to share appears to be giving way to a model where sociolinguistic data is considered to be too valuable not to share” (Kendall and Van Herk 2011: 3).

1.2 Disciplinarity and intersections

The past 50 years of sociolinguistic work have also demonstrated the great extent to which systematic variability is a pervasive and integral part of human language. As Weinreich et al. (1968) wrote, a language without variability is both nonfunctional and inconceivable. Variability in form, in structure, and in meaning allows human language its range of expressiveness, its ability “to do things” (e.g. Austin 1962[1975], Searle 1969), and, finally, its ability to change. Variation in language is the explicit focus of research in many areas of sociolinguistics, especially the VARIATIONIST tradition associated with the work of William Labov (e.g. 1966[2006], 1972) and the growing field of SOCIOPHONETICS (cf. Thomas 2002a, 2011a, Foulkes and Docherty 2006). This book grows out of these traditions, but it also seeks to be about something more. In these pages I attempt to connect work in sociolinguistics to other research paradigms in other areas of language study, in particular within psycholinguistics and social psychology. As we will see in Chapter 2, pauses, and speech timing more generally, have been most actively and productively studied by psycholinguists and social psychologists. Examining these features from a sociolinguistic perspective, but remaining sensitive to the many psycholinguistic findings about them, can aid in our fuller understanding of the nature and function of language variation.

In fact, interest in language variation and, particularly, in how social factors relate to this variability, has grown outside of sociolinguistics in recent years. For instance, work on the psychology of language and within psycholinguistics has often focused on variable features and what that variability means, but most often in terms of what variability shows about speech production on the one hand and how listeners overcome variability as a “problem” for speech perception on the other. Quite recently some of this work has begun attending to the role of subjects’ dialect and personal backgrounds more directly. In a 2009 paper published in the Journal of Memory and Language, Meghan Sumner and Arthur Samuel examined the perceptual processing of productively /r/-ful⁴ and /r/-less New Yorkers and /r/-ful non-New Yorkers and found significant differences both between non-New Yorkers and New Yorkers and between the two New York groups, despite both of the New York groups receiving similar daily exposures to the same /r/-less variants. Instead of stopping there, Sumner and Samuel went on to consider what this may mean for an understanding of “dialect,” which despite being widely acknowledged as problematic to define has always been understood (implicitly at least, if not explicitly) as a configuration of productive features of a speaker’s or group of speakers’ language. Sumner and Samuel’s results appear to indicate differences in the underlying representations of the forms for these speakers, and the authors suggest that dialects should be considered (or even defined) not only in terms of speakers’ productions, but also in terms of their perceptions and representations. They further offer that these three “aspects of a dialect may differ within an individual, just as they differ between individuals” (Sumner and Samuel 2009: 500). Other recent research (e.g. Strand and Johnson 1996, Evans and Iverson 2004, 2007, Hay, Warren, and Drager 2006, 2010, Kendall and Fridland 2012, Fridland and Kendall 2012) has examined the role of social factors on the perception of linguistic forms, but I mention the Sumner and Samuel work because it makes explicit a need for such work, and for sociolinguistic work generally, to consider more deeply its underlying assumptions about “what it means to have a dialect” (Sumner and Samuel 2009: 500) in the first place.

Nonetheless, there are of course major differences between sociolinguistics and psycholinguistics. Psycholinguistic research is most often undertaken in the laboratory, in highly controlled settings, while sociolinguistic research is most often undertaken in the field in settings and ways that might maximize the naturalness of the spoken language, that is, that minimize the OBSERVERS’ DILEMMA (cf. Labov 1972, Milroy 1987) rather than control the possible sources of variation. It is also true that the main research questions of sociolinguistics and psycholinguistics differ greatly. Yet, I believe it is fair to say that each of these fields studies variation and is interested in what that variation means. For socio linguists, interest is often in variation because it yields insight into the extralinguistic, social factors in language in use, and, for scholars who follow Labov’s variationist paradigm, in that understanding variation is central to understanding language change. For psycholinguistics, variation is often useful as a window into the processes of language production and a source of potential difficulties in language comprehension and processing. Variation pervades both of these fields and both have yielded great insight into the causes and meanings of that variation. Yet, for most of their histories, research in these fields has operated independently. To make an observation that is surely overly simplistic: sociolinguists publish in sociolinguistic journals and psycholinguists publish in psycholinguistic journals. There is just too much to read (and moreover to do) for us to follow everything of interest. Yet, to understand variation and its role in human language more fully greater collaboration is needed across these disciplines. Perhaps the time is right to pursue a more collaborative SOCIAL PSYCHOLINGUISTICS?

But a label is just a label, and, while I think this label invokes some ideas worth considering, my goal is not to dwell on terminology in these pages. Further, this book is surely not the first place to consider such a thing as a social psycholinguistics (though the collocation is surprisingly rare).⁵ As I mentioned earlier, psycholinguists and laboratory phoneticians have recently begun to pay closer attention to the literature on socially differentiated language variation (such as the work by Sumner and Samuel). The burgeoning field of sociophonetics (cf. Thomas 2011a, Di Paolo and Yaeger-Dror 2011), with its instrumental and often experimental methods, bridges some traditional gaps between these research disciplines. (Readers are referred to Thomas 2011b for a recent review of work relating sociolinguistic variation to cognition.)

So, while I write this book primarily as a sociolinguist, I see the boundaries of these two approaches – sociolinguistics and psycholinguistics – as overlapping, and ultimately, almost nonexistent. Where do social factors disappear or become irrelevant? Where do cognitive factors cease to impact language production and perception? I approach the questions of this book from the view that separating these two sets of factors within a thorough study of actual conversational speech is about as possible as imagining a language without variability.

As I wrote above, a major disciplinary difference between the importance of variation to sociolinguistics and psycholinguistics is how that variation informs our understandings of language and our theoretical perspectives on language. A second major difference has traditionally been in methodology. The field-based studies of sociolinguistics are a kind of corpus-based linguistics, with the fieldwork generating richly contextualized corpora of natural speech data.⁶ Psycholinguistics, on the other hand, has traditionally used experimentation to gather its data and test its hypotheses. Increasingly, however, these methodological differences are blurring and numerous sociolinguists have taken to lab-based, experimental methods (cf. e.g. Campbell-Kibler 2005, 2007, 2010, Hay, Drager, and Warren 2009, Drager 2010; see Thomas 2002b for a thorough and historical review). Psycholinguists have also increasingly incorporated (most often standardized) corpora and corpus analyses into their research projects (e.g. Clark and Fox Tree 2002, Bell et al. 2003, Kapatsinski 2010, just to list a few). Ultimately, I believe that both of these approaches to empirical linguistic analysis are necessary to better understand language variation, change, and processing. Nonetheless, in this book, I limit my focus to corpus-based examinations. Several of my suggestions and findings in later chapters point to the need for further experimental testing and doing so would surely strengthen the findings of this research. However, for space, time, and focus, I maintain a strictly corpus-based view here, with the aim of exploring just what we can learn from such an approach.

1.3 Why exactly speech rate and pause?

It is worth in this first chapter to ask why we might want to study speech rate and pause rather than some other features. Especially as a linguist and a sociolinguist, why should I (or you for that matter) be interested in these features, beyond the fact that they are amenable to large-scale corpus-based analysis? The answer, I believe, is as follows.

Rate of speech and pause are ubiquitous features of human language. Every utterance by every speaker of every language (even sign languages) can be characterized as having a particular rate of production and by being in relation to some intervals of silence. Further, silence in speech is a critical part of expression. A large proportion of talk in action is, in fact, silence – that is, comprised of the pauses between speakers’ utterances. For some of the source data examined in this book, as much as 35 percent of the transcribed recording is in fact silence on the part of the participants! (Admittedly, these high numbers are from particularly reticent participants and figures of about 15 percent are more typical.) By looking closely at these omnipresent phenomena we can gain insight into larger patterns of variation, and variation in less common features.

A related question then would be why do I only examine speech rate and pause. Other temporal features – e.g. segment durations and speech rhythm – are also relevant here and would also be usefully examined in the context of a large-scale corpus sociophonetic analysis. The answer here is two-...

Table of contents

• Cover
• Title Page
• Copyright
• Contents
• List of Figures
• List of Tables
• Acknowledgments
• Part I Speech Rate, Pause, and Corpus Sociophonetics
• Part II Studies in Speech Rate and Pause Variation
• Part III Speech Rate, Pause, and Sociolinguistic Variation
• Appendix I: Guide to the Website
• Appendix II: Correspondences between log-millisecond (log-ms) and millisecond (ms) pause durations
• Notes
• References
• Index