Review of Alexander Bergs

Alexander Bergs, Social Networks and Historical Sociolinguistics. Studies in Morphosyntactic Variation in the Paston Letters (1421-1503). Mouton de Gruyter [Topics in English Linguistics, 51], 2005, 318 pp. ISBN is 3-11-018310-2.

This is an ambitious and stimulating work, because the writer includes both a detailed study of three linguistic variables in the writings of the Paston Family and a contextualisation of these findings into the current state of our linguistic knowledge as a whole. The methodological issues involved in the analysis of historical data are discussed at some length and, in spite of the book’s title, the discussions are not limited to a purely social network approach. Many different areas of linguistics are referred to, many questions are raised. In contrast to the current trend in case-studies that confine themselves to a limited set of hypotheses related to narrowly selected correlation patterns and their quantitative interpretations, Bergs uses his own case studies (presented in the middle chapters) to question and illuminate a large number of methodological issues.

An outstanding feature of this volume is the large number of parameters that it accepts (individual, group, network ties, three groups of linguistic variables, three generations of writers, intra- and extra-linguistic factors). This leads to a very dense piece of research. Bergs’s admirably lucid and at times conversational style and his frequent summaries are therefore necessary and particularly welcome. The decisions he has had to make concerning descriptive background and analytic technique are clearly set out and well justified, although more dialectal information about relativisers and light verb constructions would be welcomed by this reader. Overall, the book is very readable; Bergs’s deep interest in the subject is infectious and one enjoys accompanying his thoughts as they light on different linguistic and theoretical issues.

Such an ambitious undertaking is not without its dangers, however, and after an outline of the contents of the book this review will concentrate on two areas of methodology that are especially challenged by this approach, one internal to social network analysis (the establishing of individual network strength scales), the other a general issue concerning the place of the individual in studies of social behaviour

The research presented in this book focuses on three late Middle English variables: personal pronouns, relativisers, and light verbs (complex predicates). Any researcher interested in these variables should read these analyses, because the discussion covers a wide variety of theoretical and analytical perspectives, and, importantly, discrepancies between the results of this work and the results obtained from earlier studies should now be taken into account. The methodological and theoretical contexts of the analyses are put together in what might almost be called a hidden agenda of the book, which seems to be an effort to question our techniques and refine our understanding of the processes involved in language change, as revealed by the emphasis in twice presenting a model of language change, illustrated on pages 42 and 256. Those interested in the theory and techniques of historical sociolinguistic analysis can also, then, benefit from this work.

Bergs clearly sets out his aims and concerns in the short Introduction: the Paston letters were chosen for this study of language change because of the particularly interesting period in which they were written; personal pronouns, relativisers and light verb constructions were chosen because all were in transitional phases at that time and there are unresolved issues in our understanding of each of these changes; and the social network model was chosen as a tool for analysis because the nature of the source materials (letters from family members) and the subjects of enquiry (which include, crucially, the role of the individual in change) are amenable to this sort of analysis.

Chapters 2 and 3 discuss “Historical Sociolinguistics” and “Social Network Analysis”, respectively. After a wide-ranging survey of issues concerning the relationships between linguistics, history and the other social sciences, some of the main issues underlying the study of historical sociolinguistics are discussed. These range from questions concerning the reconstruction of the language and meanings of long-dead speakers out of written records, to style, register and grammaticalisation.

Chapter 3 (“Social Network Analysis – present and past”) presents an overview of social network analysis in general and in the study of language variation and change in particular. Historical network analysis is then scrutinized, again from the viewpoint of historical sociolinguistics. Then, “an attempt [is] made at developing some general principles and techniques for social network analysis in (late) Medieval England, and at analyzing the networks of the Paston family with such instruments” (p. 22). Finally, Bergs describes his materials in detail. In this chapter, ideas from different theorists are brought together in a useful discussion of similar and overlapping concerns. Labovian “leaders” are equated with Milrovian “early adopters” (p. 39), for example, and a result of such syntheses is a detailed but clear diagram of “innovation type and change in networks” (Fig. 8, p. 42).

The following three chapters, Chapters 4, 5, and 6, deal separately with the descriptions and analyses of the selected linguistic variables. Each chapter provides a condensed history of the variable under investigation and a review of its literature. The discussions are extended into other areas of linguistic enquiry, with links drawn between, and questions asked about, what has already been found, what remains to be discovered, and how these existing and future findings may fit into theories of language from Generative Grammar to Cognitive Linguistics. Beyond this, however, generalizations cannot be made, because, as a major tenet of the volume borne out in practice, each variable requires, and attention is accordingly given to, different developmental and usage factors. Thus, for instance, while all three analyses pay attention to the social variable gender, for the personal pronouns the roles of dialect and linguistic analogy are discussed, while these have no place in the analyses of the following two variables; and again, while the relationship between the author and the addressee is considered in analyses of personal pronouns and relativisers, it is not a part of the analysis of the light verb constructions.

A few selected examples of findings, both positive and negative, from these chapters follow: in Chapter 4 it is, interestingly, discovered that the marking of thou/you alternants for social relations has not entered the written language of these people; at the same time no “communal patterns in pronoun usage” are found (p. 127). An important finding in this chapter is the clarity of the generational shifts in the Pastons’ usage of hem/them and here/their (p. 106) and the subsequent observation that an interpretation of this as a clear case of generational shift needs to be modified in view of the fact that the same individual (officially from the same generation, therefore) can in fact change his or her usage significantly during his or her lifetime. The outstanding example here is Margaret Paston; of the “dramatic” changes she makes in her usage Bergs notes that “the generational shift ... must ... be taken cum grano salis, as these presuppose more or less stable language use throughout the lifetime of individual speakers” (p. 113).

In Chapter 5 the postulated differential treatment of restrictive and non-restrictive relative clauses is supported by the material, and animacy of the antecedent is shown already to be important in non-restrictive clauses, but the relative frequencies of which and the which are contrary to those posited by Mustanoja (cited on p. 164) and found in the Helsinki Corpus (p. 165). Finally, in Chapter 6 a hierarchy of (increasing) markedness for light verbs is identified as give>have>take>do>make (p. 232). It may be of interest to compare this result with the frequency rankings presented in Tanabe’s study of these verbs in the Paston letters (Tanabe 1999:101, 130), while noting that the two researchers concentrate on different selections of both the verbs and the writers. Contrary to the expectations of earlier studies, and to a general (if weak overall) trend of Tanabe’s decade-by decade analysis, no linear increase in the frequency of light verbs is found in Bergs's corpus, instead it is found that for one of the three generations studied (within the years 1451-1475) “the more often speakers used light verb constructions, the fewer different types of light verb constructions (i.e. different nouns) they used, and vice versa” (p. 237). This generation, by the way, coincides with the two decades showing the highest frequencies of use in Tanabe’s study.

With very many subjects broached in the individual analytic chapters, the Conclusion (Chapter 7) has many strands to tie up; it cannot merely recap a previous hypothesis and point to satisfactory results, because the study asked many open questions and the results of many of the analyses were negative. That is, important questions concerning issues such as the role of marking and saliency in the spread of change, the extent to which choice of variable is conscious, and the extent to which the individual language users are involved in language change, with all their theoretical and methodological repercussions, run through the book and cannot be simply concluded. Further, in spite of Bergs's optimistic expectations for this sort of analytic approach to late Medieval English as expressed in the conclusion to his earlier work on such material (Bergs 2000:251), it was found that applying Social Network analytic tools to this corpus did not offer clear and illuminating answers to all of the questions, but presented, rather, a mixed bag of results that cannot be presented in a simple and unified whole.

This is not to say that the work is inconclusive, however. Many valuable results have come out of the study and most of them are mentioned in this last chapter. One, for instance, is the fact that different linguistic variables were shown (as predicted in the introduction) to require different methods of analysis and interpretation. This brings the issue of methodology to the fore once more, not that it has ever receded very far into the background. It seems counter-logical to seek a single and consistent analytic framework that at the same time treats differently each speaker, variable and parameter. This sort of dividing up of the analytic work is further complicated by the apt observation that there is a qualitative difference between the usages of a first-generation user of a certain form, who is in some way breaking new ground, and the following generations who, while they may show a statistically smooth incrementation of usage, are nevertheless involved in quite a different sort of choice of forms, there being by now an existing repertoire for them to pick from (p. 245).

Bergs notes that further consideration must also be given to the role of age grading in language variation and change, and the concomitant question of how to account for it methodologically. Indeed, the whole issue of change on the individual level – changing social and personal circumstances, changing network ties, changing language usage – is shown to be a major obstacle to effective social network analysis of historical material that spans a “movie shot” rather than a “snap shot” of time (p. 260).

In Bergs's analytic chapters it turns out that only some individuals showed very clear patterns of linguistic behaviour that could convincingly be explained in terms of their biographies or positions within the network (notably Margaret [traditional usage], Elizabeth [innovative, a social climber] and the elder John the third [most advanced user of the variables; travelled around England and spent time at court]). For the Paston family network as a whole, even within the analyses of single variables, this was not usually the case. When the results of the three variables were viewed together with the individuals’ network strength scales, “no uniform correlation pattern” could be found (p. 254). In the end, Bergs is forced to conclude that the network strength scale in this research “only represents average network structures across life times” (p. 258) and can serve his research only as “orienting statements” (p. 262).

I suspect that a comment Bergs makes earlier (pp. 54-55), concerning the differing roles of networks in changes where there is an accepted standard variety and in changes where there is none, may be a more powerful explanation than he credits for his material’s lack of overall correlation between network strength scores and form frequencies. Although network analyses are not reliant on the fairly rigid class models that stratificational studies use, it is possible that the nature of many of the ties selected for scoring may in fact be geared towards correlations between individuals’ social positions and (linguistic) behaviour that can be placed on some sort of normative or even prestige scale. For while power relations and education, to choose but two, may relate in a particular society and time to a greater diversification of social ties, they are also directly related to position vis-à-vis the standard or prestige usage where either exists. But not all language variables have this sort of social meaning or indeed this sort of hierarchical relationship between their different forms. This does not mean that one has to reintroduce that questionable concept-of-convenience the “free variable”, it means only that what a variable correlates with does not have to be a crudely socially evaluative factor. This, by the way, fits in obliquely with another preoccupation of Bergs, that of the roles of saliency and markedness in the individual’s choice of variable form. His discussion in the conclusion (pp. 255-257) could perhaps be expanded beyond the issues of salience and conscious language use, to include this insight, too.

As Bergs may be lightly touching upon (p. 20), and Singh (1996a:8) certainly insists upon more firmly, what a variable form correlates with could be a matter of any sort of meaning in the broader sense. The failure of the particular coordinates chosen in Bergs’s Network Strength Scale to show a meaningful pattern may indicate that, contrary to what he reported early on in the book in – to be fair – a different context, it is not a case of “for the sociolinguist ... any kind of variation will do” (p. 18). Not only the social situation but also the linguistic situation (lack of a recognised standard in this case) must be matched by adjustments to the analytic toolbox (here the selection of scored elements), and each society and historical period has different sets of factors relevant to such a study. It is increasingly being recognized that no single category (e.g. “gender”) is universally applicable as a significant social variable; furthermore, categories may themselves contain sub-categories that need to be taken into consideration. To continue with the example of the category “gender”, we may note that in some research it has been found to act as a scalar value rather than as a binary distinction. In addition, along with other issues relevant to the issue of network ties and scoring, Fitzmaurice (2000) has noted that “interpersonal ties differ in kind between men and women as well as in strength” (2000:270), a fact that, along with other non-reciprocal relations, may complicate any attempt to develop strength scores representative to some degree of the networks they claim to describe. The more we study these issues, the greater becomes the “challenge … to show which ties are meaningful to the groups and the individuals who are being studied” (Milroy 2000:220).

In retrospect, then, I wonder if Bergs still stands by the network strength scoring system he used. This is a personal comment of mine and not substantiated as far as I know by any research, but I wonder about three items in particular: firstly, I could not understand how a stay of more than (just) one week can be considered long enough to be a criterion for a “place of living” that in its turn effects the potential number of network ties a person has (p. 73). In fact I find this so odd that I wonder if it may be a misprint. Secondly, I am not sure of the usefulness of scoring the married/single criterion for this material. It seems to me that all married people were once single, often become single again, and on marriage do not necessarily reduce their number of ties; in fact, the assumption that “marriage meant generally fewer ties” (p. 73) does not seem justified to me at all, neither for men nor for women. At any rate, from the material with which I am more familiar – coming from two centuries later - we can find very many conventional individuals for whom marriage was no reducer of ties, the contrary could even be argued. Indeed, for “my” women of the seventeenth century (I’m thinking here chiefly of the many women whose letters are found amid the Verney papers) it could easily be said that marriage greatly increased the number and variety of their ties. Furthermore, the fact that a criterion like single/married, whose effects in terms of ties must surely be at least open to doubt, is given the same numerical weight as (formally) educated/uneducated could also benefit from further discussion. This leads to the third point, which is the inclusion of both gender and education as scored factors. As Tieken-Boon van Ostade has noted, “in older stages in the history of English, the variable gender ... is often closely tied up with education: men are usually educated, while women are not” (2000:214), and this is the case for Bergs's materials as well. So women receive one density point for being women and a further two for having received no education (p. 72). Might this not result in a distorted distinction between the men and the women?

Undoubtedly Bergs is concerned with, indeed at times almost haunted by, the seemingly unresolvable gap between the individual and the social or communal as it effects studies of language variation and change and, indeed, all social sciences. In each chapter of his book there is a reasoned movement between community and individual, with correlations between language usage and generation, or network strength scale being only a part of the analyses used, and then only for particular subsections of the analyses.

A search for introducers of change and “early adopters” is, of course, part of an attempt to identify bridges between smaller and larger units: between the individual and his or her network, between networks and wider communities (including “communities of practice”, for which see Eckert 2000, discussed in Milroy and Gordon 2003:118ff.), and between these communities and society at large; but because network analyses do not usually look outwards to the broader societies in which the individual networks are embedded, they remain at the micro-level of investigation and can only tentatively indicate bridges that work at the theoretical and macro levels of language change in general. On the positive side, though, “designating the relationship between a given set of individuals as a focus for study sidesteps the hazard posed by an incomplete knowledge of the whole social group in which the set operates” (Fitzmaurice 2000:274). Each analytic method has its advantages and its limitations, and some of the questions that Bergs asks knock against the boundaries of his chosen technique. As Watts (2003) has shown in his study of English borrowings in Swiss-German dialects, a network approach can over-simplify the relation between social networks and the actuation of language change; networks (or communities of practice) involved in change can be – perhaps often are – “enormously complex, involving close-knit as well as open networks, overlapping and interrelated networks and even networks that appear to have been mutually antagonistic” (2003:116).

Romaine (1996:100) has analysed the individualist vs. collectivist problem succinctly as one of two distinct levels of abstraction and, in practical terms, methodological decisions respecting this distinction, especially in matching techniques of analysis with aims of investigation, have provided a way out of the dilemma. The cost is, naturally enough, that any set of results is only illuminating relative to its theoretical and methodological origins. It is argued in Bergs's book that large-scale corpus studies “often lead to misconstrued images of actual language use in individual speakers” (p. 6), and although anyone who has read Labov’s Philadelphia work will agree that not all huge projects ignore the individual speaker and his or her importance in language change, it is generally true that a widely based corpus accounts for neither the full effects of the change upon the individual user nor the individual’s full effects upon the change; and while a network study of a small group or coterie, such as Fitzmaurice’s or Tieken-Boon van Ostade’s work on eighteenth-century groups, can show the sociolinguistic relationships between individuals in that group, it can not (nor is it intended to) account fully for those groups’ and individuals’ lasting effects upon the wider language community or even for the wider language community’s effects upon them. There is a sense that something is lost in the gap between the two focuses, something that relates to that methodologically inadmissible and largely undefinable factor “what happens in real life”. Maybe this is because descriptions and correlations are not explanations (Nevalainen and Raumolin-Brunberg 2003:19), and we so much thirst for explanations that we wish to read our correlations as causal. As Nevalainen and Raumolin-Brunberg put it, “it is the general quest for a sociolinguistic theory that is at issue here” (2003:19).

Putting aside as a tempting diversion the different arguments in this area, what seems relevant to the book under consideration is Singh’s call for a sociolinguistic theory where “language structure and language use will be a differentiated unity and not merely the two autonomies of the orderliness of competence and the anarchy of performance, and the theory of discourse a rational reconstruction of the actualization of discourse potentials” (Singh 1996a:2). Bergs’s underlying if loose affinity with Singh is signalled by his use of the word “free”, when he wonders “whether speakers are essentially free to choose and may do what they want” (p. 263). Singh has asked for a theory that allows the reconstruction of language as a social activity involving “joy, truth and freedom” (1996a:2), and desires an outlook that can “reintegrate, at the level of analysis and theory, what was deliberately or inadvertently left out in [the] initial search for special frameworks” (p. 4). What Bergs is implying when he asks “How much are speakers constrained by their linguistic system, how much do they actually shape this system?” (p. 4), although ostensibly expressing a desire for more attention to be paid in interpretations of data to the level of individual speaker, may in this context be seen as an unconscious plea for just such a sociolinguistics. Indeed, the whole of Singh’s argument is an appropriate background to the themes underlying many of Berg’s questions in this volume.

Having asked these questions, however, and shown some shortcomings of even a very carefully-prepared Social Network analysis in sociolinguistics, what practical solutions does Berg present? First of all it should be noted that Bergs is not attempting to constitute a new critical sociolinguistics as posited by Singh, but rather to show that “any claim about cognitive, universal, or typological determinants of linguistic change need not only hold for the level of the speech community or its subgroups, but also for a substantial number of speakers in isolation, if it wants to reflect reality” and that “variation on the level of individual speaker ... is also guided by a number of both intra- and extralinguistic factors” (p. 5). In fact, as his group correlations break down, the emphasis on interpretation of each individual’s results becomes increasingly important. Then we may see the analyses in this volume as a brave attempt to try something new while at the same time clinging to the wreckage of a favourite methodology. As noted above, the social network analysis of language provides the researcher with freedom to develop his or her own parameters of investigation, specifically in the definition of what constitutes a network tie and what score to give to each tie (Milroy and Gordon 2003:121; Milroy 2000:220), although such adaptations are not widely discussed in the present work.

What Bergs seems to be promoting in terms of analytic technique, as a way to allow that freedom a chance to show itself, is to treat each variable and each informant (both individual and community) separately and then together, both in contrast and cumulatively. It is a huge and complicated undertaking, even with a relatively small number of writers in the core group. After reading this book one may ask if the task is not, perhaps, too massive, especially when different periods of time need to be taken into consideration as well. Dealing separately with “as many factors as possible” (p. 261), and then interpreting the results “in a less empirical and more hermeneutic way” (p. 261), makes sense at the micro level of interpretation but does not resolve the problem of how, when so many factors are under consideration, to integrate the specific results with (a possibly large number of) macro level models. Quantitatively, if such an approach is allowed, it seems almost impossible; it would take an extremely ambitious multivariate analysis to attempt such a complex data set of unfixed variables and parameters. In a study of the role of the individual in language change, Raumolin-Brunberg (forthc.) concludes that the language of individuals can be studied in a historical context but, importantly, she suggests that “this is only possible if there is sufficient baseline data against which the individual usage can be compared and analysed”[2]. Such baseline data can provide the fixed parameters required by analysis. Her strong suggestion, and one that fits in with the research of the Helsinki group as mentioned in various of their publications, is that this baseline data and analysis should come from electronic corpora.

That Bergs has looked at his materials in such an unusual amount of detail, and then related his readings and his findings to a variety of different linguistic theories, is a great strength of the book, and the questions that are raised needed to be raised. While no work can provide a definitive list of elements needed for another person’s research, the total of linguistic factors considered in these chapters would certainly make a useful check-list for anyone wishing to contextualise their interpretation of similar language variables. To give an example, when investigating the possible reasons and mechanisms of change from h- to th- plural personal pronouns, Bergs considers (among other things) both therapeutic and prophylactic reasons for change (pp. 92-93), Pike’s “formatives” (p. 98) and thence the cognitive abstractions of the Wickelphone and its descendent the Wickelfeature (pp. 98-100). It is a pleasure to see all these different perspectives being made to work together towards a deeper understanding of the mechanisms of language change.

Margaret J.-M. Sönmez, Department of Foreign Language Education, Middle East Technical University, Ankara, Turkey. (Contact the reviewer.)

Bergs, Alexander T. 2000. “Social networks in pre-1500 Britain: Problems, prospects, examples”. In: Ingrid Tieken-Boon van Ostade, Terttu Nevalainen and Luisella Caon (eds.), Social Network Analysis and the History of English. Special issue of the European Journal of English Studies (EJES) 4.3, 239-251.

Eckert, Penelope. 2000. Linguistic Variation as Social Practice. Oxford: Blackwell.

Labov, William. 2001. Principles of Linguistic Change. Vol. 2. Social Factors. Oxford: Blackwell.

Milroy, Lesley. 200. “Social network analysis and language change: Introduction”. In: Ingrid Tieken-Boon van Ostade, Terttu Nevalainen and Luisella Caon (eds.), Social Network Analysis and the History of English. Special issue of the European Journal of English Studies (EJES) 4.3, 217-223.

Milroy, Lesley and Matthew Gordon. 2003. Sociolinguistics. Method and Interpretation. Oxford: Blackwell.

Mustanoja, Tauno F. 1960. A Middle English Syntax. Helsinki: Société Néophilologique.

Nevalainen, Terttu and Helena Raumolin-Brunberg. 2003. Historical Sociolinguistics. London: Longman.

Raumolin-Brunberg, Helena (forthc.). “Leaders of linguistic change in Early Modern England”. In: Roberta Facchinetti and Matti Rissanen (eds.), English Studies and Corpus Linguistics. Linguistic Insights. Frankfurt a/M: Peter Lang.

Romaine, Suzanne. 1996. “The status of sociological models and categories in explaining language variation”. In: Singh (ed.) 1996b, 99-114.

Singh Rajendra (ed.). 1996b. Towards a Critical Sociolinguistics. Amsterdam/Philadelphia: Benjamins.

Tanabe, Harumi. 1999. “Composite predicates and phrasal verbs in the Paston letters”. In: Laurel J. Brinton and Minoji Akimoto (eds). Collocational Idiomatic Aspects of Composite Predicates in the History of English. Amsterdam/Philadelphia: Benjamins. 97-132.

Tieken Boon van Ostade, Ingrid. 2000. “Social Network Analysis and the History of English”. In: Ingrid Tieken-Boon van Ostade, Terttu Nevalainen and Luisella Caon (eds.), Social Network Analysis and the History of English. Special issue of the European Journal of English Studies (EJES) 4.3, 211-216.

Watts, Richard J. 2003. “Why fuude is not ‘food’ and tschegge is not ‘check’. A new look at the actuation problem”. In: David Britain (ed.), Social Dialectology: In Honour of Peter Trudgill. Amsterdam/Philadelphia: Benjamins. 115-129.

[1] This is a revised and expanded version of LINGUIST List review number 16.2367, published on 11th August 2005.

[2] I am grateful to Helena Raumolin-Brunberg for letting me see and quote from this work before its publication.