Woodbridge, Boydell & Brewer, 2020, ISBN: 9780861933549; 362pp.; Price: £50.00
St Hugh’s College, Oxford
Date accessed: 22 June, 2021
Luke Blaxill’s book deserves to be seminal. Its unassuming title conceals a bracing methodological challenge: an argument for the application of specific digital techniques to the study of electoral politics. It deploys ‘corpus linguistics’—the computerised compilation and interrogation of massive databases of millions of words—to intervene in a series of debates about the language of the platform in late-Victorian and Edwardian Britain. And it does so with enormous conviction. The War of Words builds its case so patiently, carefully, and politely, that even the most hardened traditionalists may find themselves struggling to dig up objections. Certainly, the public language of parliamentary candidates is a genre peculiarly suited to the analytical tools Blaxill deploys, and work on other spheres of politics may not be able to benefit from the sustained application of ‘text-mining’ methods to the same extent. But the book offers an object lesson in how to present an argument about method.
The War of Words has been trailed in a series of journal articles and book chapters over the last decade, so some readers will already be familiar with the techniques involved, and a number of the more striking reinterpretations the book offers. But having everything presented together—and supported by an imposing array of statistical tables (36) and figures (44), a ‘technical glossary’, and 68 pages of methodological and statistical appendices—lends the research a powerful cumulative effect. As this suggests, the volume’s embrace of digital methods goes well beyond the searches for ‘hits’ on particular keywords, and the Google Ngrams, which are already part of the historian’s armoury. What does the book mean to do?
Its ambitions lie on three levels. In the broadest terms, it offers an ‘applied methodological experiment’ (p. 43). Blaxill suggests that, thanks to the apparent failure of early efforts at quantification in history in the 1960s and the 1970s, historians have since treated the approach with equal measures of fear and disdain. Because most of today’s practitioners of ‘digital humanities’ have chosen ‘voluntary marginalisation’, preferring to speak to the methodologically converted rather than address communities of the unregenerate, members of the historical profession have had little reason to re-evaluate their prejudices (p. 239). For all the optimistic rumination that scholars have offered on the interpretative possibilities opened up by using computers to study massive bodies of texts, moreover, few of them have actually offered the labour-intensive, practical case studies which are clearly necessary to ground such claims. This is where Blaxill steps in, aiming to show that a ‘big data’ method can act as a check on imperfect intuition, and provide better evidence and analytic precision.
The book’s second set of goals relate to modern British political history in general. Echoing various other historiographical commentators writing in recent decades, Blaxill find that the turn towards the description of political-cultural phenomena (including language) has gone too far, and calls for a return to problems of causation, alongside closer consideration of the connections between spheres of politics too often held apart from one another. He proposes that the method adopted in his book ‘can help overcome the challenge of reintegration’ (p. 9).
Finally, The War of Words turns towards more specific problems in the study of late-Victorian and Edwardian elections and electoral culture. The meat of the volume consists of an analysis of the language used by parliamentary candidates in the nine general elections held between 1880 and 1910.(1) The evidence base is composed of five million words of stump speeches, as reproduced in the local and national presses. Blaxill has compiled three main searchable ‘corpora’ of texts , the first a local set from East Anglia (Norfolk and Suffolk); the second containing the speeches of frontbenchers; and the third sampling speeches made in constituencies of diverse characters from across the country, varying from election to election. As this suggests, significant efforts have been made to cover the bases.
The book’s first chapter is historiographical, theoretical, and methodological—making a general case for corpus linguistics and text-mining in history—and can profitably be read even by those without much interest in elections, or modern Britain. The other four chapters work through the elections of 1880-1910 two or three at a time, with reference to particular issues in the rhetorical campaigns which surrounded them. Chapter 2 is about the impact of electoral reform, and Joseph Chamberlain’s ‘Unauthorised Programme’, on the language of the 1880 and 1885 elections; Chapter 3 deals with the place of Home Rule in the campaigns of 1886 and 1892; Chapter 4 focuses on the idea of ‘imperialism’ among the candidates of 1895 and 1900; and Chapter 5 covers the impact of ‘New Liberalism’ and social reform in 1906 and in the two elections of 1910. The book reflects on longer-term trends throughout.
The marriage between chronological and thematic approaches is not always entirely happy, as when a (valuable) discussion of Liberal ‘faddism’ is tacked on to the end of the chapter on empire. But for the most part it works, because The War of Words does not aim to sketch out any kind of grand, general reinterpretation of the content of electoral language. Its purpose is to vindicate a method, and to that end, every chapter is designed to intervene in one or more specific historiographical debates. This is not just a question of topping and tailing: at every turn, Blaxill is at pains to trace his point-by-point agreements and dissents from the existing literature on the campaigns he discusses. The odd longueur here is easy to forgive, because this sustained, deliberate engagement is so fundamental to what the book is trying to do.
Readers may by now be wondering what the main body of the text actually looks like. Boiled down, the book is about patterns of word use. It is about how and how often specific groups of words, connected with particular political concepts, issues, and personalities, were deployed by different groups of candidates for election to parliament. As well as furnishing raw numbers, Blaxill’s software also allows him to track the ‘lexical attraction’ between terms, i.e. how likely they are to appear in proximity to one another within a corpus or corpora, alongside a number of other pieces of technical wizardry. But the book offers much more than just counts and ratios. Keywords are regularly placed in their immediate discursive contexts, in order that the different or changing meanings of multivalent terms—like ‘class’ (pp. 191-8) —can be understood and integrated into the analysis. So Blaxill is sensitive to the texture as well as the shape of electoral language, and the book is full of quotations from the speeches of its parliamentary candidates. All this helps to make its arguments more accessible, comprehensive, and persuasive.
Blaxill’s priority, however, is to generalise more reliably. A salutary and rather sobering passage in the first chapter highlights the shakiness of the foundations on which broad claims about popular politics by good historians often rest (pp. 32-3). The War of Words suggests that by assembling giant corpora of speeches, and applying the proper statistical techniques needed to approximate representativeness, it is possible to make more robust arguments about what did and did not matter in election campaigns. Some of the results are incredibly striking. The chapter on the place of ‘imperialism’ in electoral language is a case in point. Blaxill systematically takes apart the (large) literature which argues that empire was a leading issue in elections throughout this period, suggesting that its significance was only intermittent; and shows also that, against the idea that imperialism was a ‘contested discourse’, Conservative candidates talked about it far more often than Liberals did. Richard Price’s well-known thesis on the ‘khaki election’ of 1900 is ‘entirely rejected’ (p. 144), and assuming Blaxill’s numbers are reliable, the landscape around two major historiographical problems looks quite different.
Beyond ‘submitting its methodology to trial’ (p. 39), the book in its conclusion also makes some important general points about electoral language and culture. These rest more on qualitative than quantitative evidence, and should be taken seriously even by historians who remain suspicious of Blaxill’s technological predilections. The first is about the ‘elasticity and ephemerality’ of platform speeches (p. 229). The War of Words suggests that issues and personalities came and went rapidly between election campaigns, that and topicality was so prized in the cut and thrust of electoral debate that party traditions and political theories were less present on the platform than historians have imagined. Blaxill’s conclusion here is that ‘electoral language should be regarded as a distinct discourse where issues, values, traditions, personalities and ideologies could play out differently from in other spheres of politics’ (p. 230). He hints, further, that election speeches may even have helped to shape wider party thought, strategy, and doctrine. His second main general point is that the platforms of rival political parties were closely interrelated. Issues emphasised by one side were nearly always responded to by the other, such that we should see politicians fighting elections as ‘more debaters than they were speakers—men whose charge was to propose, oppose, rebut and ultimately out-argue their opponent’ (p. 231). As such, Blaxill casts doubt on the wisdom of studying political parties in isolation, stressing the ‘interactivity’ of speaking campaigns, and the consequent value in ‘parallel readings’ of parties’ languages (p. 232). Here he echoes a point made by the original historians of the so-called ‘high politics’ school writing in the 1960s and 1970s, and developed by their inheritors, but one which needs reiterating when it comes to the study of electoral politics. Finally, the book makes the case that—on balance—the Conservatives were more successful than the Liberals in adapting to the conditions of late-Victorian and Edwardian platform politics, thanks to their message discipline and unremitting partisanship, though the Liberals did often set the terms of debate (pp. 233-8).
The book works as well as it does because it is so unyieldingly reasonable. There is no trace here of the Young Turk excitably advocating for the Next Big Thing: Blaxill undertakes his task of selling a new method to a sceptical audience with considerable skill. He accepts that the approach being introduced is controversial. But he patiently anticipates and meets as many objections as he can, in passages that appear to bear the impress of a number of hostile seminars. Much potential criticism is disarmed, moreover, by the fact that the book argues not for a new paradigm in which corpora are king, but for ‘a middle path… between a nuanced humanistic reading and a more purely linguistic, empirical, scientific one’. Blaxill suggests that it is in the ‘friction’ between close reading and computerised analysis (or ‘distant reading’) that the value of the latter might come most clearly into focus; that text-mining is something that can be ‘dabbled with’ alongside established methods; and that it can ‘augment – and ultimately empower – traditional approaches’ (pp. 13-15, 239). He emphasises that text-mining should be used ‘practically’ to re-examine consensuses (p. 20). And it is hard to object to his point that even if corpus analysis is employed only for ‘initial textual reconnaissance’, it could bring to light issues which might otherwise have gone unremarked, and that even then ‘we may still reflect that we have discovered a rather powerful new tool’ (pp. 35-6, 43).
Taken on its own terms, it is not easy to find fault with the book, which is a testament to how carefully thought-through and precisely written it is. The War of Words is unquestionably a path-breaking study, and it will be fascinating to see how attractive the ground it has cleared proves for other scholars. Zooming out a little, however, there are a few questions we might ask about the theory and practice of the project Blaxill has undertaken.
Verifiability is one obvious issue. The book does not enumerate the contents of any of the corpora it uses—doubtless the idea of doing so would have given the publisher conniptions—but it could perhaps have been a little more expansive in discussing the principles by which they were compiled. For instance, who precisely counted as a ‘frontbench politician’ for the ‘National’ corpus? Did under-secretaries? Did Gladstone, after he surrendered the Liberal leadership? Some historians will take issue with the exclusion of Liberal Unionists from this corpus until 1895, and their inclusion with the Conservatives thereafter, but at least that decision is clearly stated (pp. 243-4). There are some further unanswered questions, however, about how the ‘Constituencies’ corpus was assembled.(2) So Blaxill’s argument that computerised history involves giving readers privileged access to the ‘basis and parameters’ of the historian’s judgement, thanks to its necessary ‘democratic’ transparency, is not entirely borne out in his own work (pp. 34-5). But this criticism could easily be met in future studies.
The other, more fundamental issue here is that most historians lack the expertise to judge Blaxill’s technical work. It is to be hoped that other reviewers will be better able to deal with these aspects of the book, as this one certainly cannot. That Blaxill is mainly addressing a readership of Luddites is not, of course, his fault. He does his best to bring them into the fold, stressing that text-mining can be ‘deceptively simple’ (p. 239). But his conclusions would likely demand modification, at least, if it could be demonstrated that the text of his corpora were less reliable than he claims, that the sampling or weighting techniques he has employed were faulty, or indeed that bugs in the software somehow affected the outcome of the queries he ran. It is difficult to imagine that the scientific standards of peer review to which work like this might ideally be held, including the replication of results, will evolve in the historical profession any time soon. But the point is that something more has to be taken on trust with quantitative work of this kind than it does with more ‘traditional’ scholarship, given the practical inaccessibility and unintelligibility of much of the evidence.
Blaxill here faces the problems of the pioneer. It may be that in ten or fifteen years’ time we will all have access to the same massive, comprehensive, open-access corpora, assembled according to agreed standards, in which case these issues will evaporate. But while scholars who want to engage in text-mining still have to do the painstaking work of compiling their own databanks, such question marks will persist. Disciplinary conservatism may here prove more powerful than methodological opposition: not many historians enter the profession in the hope of spending their days operating databases. But perhaps we will all be surprised.
As far as its contribution to modern British political history is concerned, there are a few more specific questions we might ask of the book. The first is about its approach to party. The book is very much about Liberals versus Conservatives (with smatterings of Labour and Unionists where appropriate). This is the central dividing line that runs through the whole statistical appendix, and the majority of the figures in the text. Certainly, most candidates for election to parliament after 1880 had to array themselves under one of these two banners, and the book unquestionably shows that concentrating on party as the main organising category for its parliamentary candidates can produce handsome interpretative returns. Its readers may be left wondering, however, about different ways of cutting the cake. As Blaxill knows, late-Victorian and Edwardian political history was as much about battles within as between the major parties: his text reflects on many of these divisions, especially inside the Liberal Party, and there is some discussion of how patterns of speech varied between different types of constituency. But what might be revealed if other dividing lines were prioritised more systematically, and party relegated to second place? To take just a few of the categories which historians have seen as relevant to the politics of this era, queries might plausibly be run on the language of parliamentary candidates by their religion, age, wealth, education, financial interests, locale, imperial connections, and (for some) theoretical commitments. Clearly, many of these divisions correlated strongly with party alignment, and it may well be that asking such questions simply confirms that party was indeed the most salient category for understanding the language of electoral politics. But it might be that other, more unexpected, inter-party linguistic patterns are revealed, and the map of late-Victorian and Edwardian politics has to be redrawn. One wonders also how far there might be anticipations of the Liberal split over Home Rule in the subtleties of the public language of the party’s candidates in 1880. All this extra work, however, might easily require another book.
Second, The War of Words is surprisingly diffident about how (or whether) the words uttered in election campaigns actually mattered. This is really to Blaxill’s credit as a cautious, sceptical analyst of the evidence. But it does raise questions. The book does not draw any distinction between candidates who won and candidates who were beaten, so it is not possible to see whether advocating particular issues or emphasising certain themes correlated at all with getting elected. And in leaving it an open question whether ‘the thousands of words issued from the platform really did count for something at the polls’ (p. 238), we are slightly left hanging when it comes to the resolution of Blaxill’s ‘explanation’ and ‘reintegration’ agenda. Elections in the late nineteenth and early twentieth centuries were not, of course, fought only from the platform: as historians including James Thompson and Kathryn Rix have recently reminded us, they also involved rich visual cultures, and impressively sophisticated forms of targeted communication with voters.(3) Many of the sources generated by these other kinds of campaigning are simply not amenable to ‘big data’ analysis, so the task of interrelating Blaxill’s findings with those of this wider literature will be challenging. The understanding of platform language which Blaxill has facilitated here is a significant advance, but it is not yet clear how far it helps us to understand other electoral phenomena.
Election speeches are clearly an ideal source-base for Blaxill’s method: copious, purely textual, generically similar, comparable over fixed points in time, and rhetorically charged. With modifications to the approach it is possible to use it to study parliamentary speech, as Blaxill has started to do for the twentieth century.(4) Other historians would certainly be able to apply the method to the language of elite and popular periodicals, and newspaper commentary, in Victorian Britain. What is most exciting, however, is less the prospect of a new wave of semi-‘pure’ text-mining studies, but more the potential that corpus linguistics clearly holds—on this evidence—to highlight unexpected patterns, and to open new interpretative doors, for political historians of all kinds.
Blaxill’s justification for performing his methodological experiment on late-Victorian and Edwardian elections—that the historiography is mature, and debate within it has been fierce—is a sound one. But it is noticeable that nearly all the literature with which his book is most closely engaged is at least 20 years old, and in many cases considerably more venerable. In the historiographical marketplace of 2020, jammed with work on global, transnational, and connected histories, and preoccupied with time, space, race, and identity, the volume might seem easy to overlook. But that would be a mistake, because The War of Words is a ground-breaking study. Its attempt to domesticate the apparently fearsome field of computational linguistics deserves to reach a wide audience, and modern British political historians are lucky that it has landed in their field. Technologically-assisted ‘distant reading’ is, on this showing, something that many of us ought to start taking more seriously.
- Not eleven, as the book suggests on p. 17.Back to (1)
- For the composition of the ‘Constituencies’ corpus, the largest of the three main corpora at 1.8 million words, we are referred back to Blaxill’s doctoral thesis, at pp. 321-5. Looking at the version of the thesis available on the King’s College London website, however (https://kclpure.kcl.ac.uk/portal/files/103826192/2012_Blaxill_Luke_04285...), the relevant appendix appears to be the one at pp. 227-9, and this only goes as far as 1900. The principles of selection are also not explained any further beyond the claim that the seats chosen are ‘roughly representative’ of Neal Blewett’s typology of British constituencies.Back to (2)
- James Thompson, ‘“Pictorial Lies”? Posters and Politics in Britain c. 1880-1914’, Past and Present, 197 (2007), pp. 177-210; Kathryn Rix, Parties, Agents and Electoral Culture in England, 1880-1910 (Woodbridge, 2016).Back to (3)
- Luke Blaxill and Kaspar Beelen, ‘A Feminized Language of Democracy? The Representation of Women at Westminster since 1945’, Twentieth Century British History, 27 (2016), pp. 412-49.Back to (4)