The aim of Change in contemporary English: A grammatical study (hereafter CCE) is to give an account of how ‘grammatical’ (i.e. morphosyntactic) aspects of English have been changing since the mid twentieth century, with some look-backs to the 1930s. The focus is on similarities and differences between written standard British and American English (BrE and AmE, respectively). The study is based on a rigorously careful analysis of comparable electronic corpora, and on a ‘bottom-up’ corpus linguistics methodology that works from ‘the identification and quantification of formal phenomena in texts towards the functional interpretation of the resulting findings’ (p. 236).

The corpora that are the basis of the study are the ‘Brown family’: the Brown Corpus (1961, AmE), the Lancaster–Oslo/Bergen Corpus (1961 LOB, BrE,) and their updates, the Freiburg–Brown Corpus (1992 Frown, AmE) and the Freiburg–Lancaster–Oslo/Bergen Corpus (1991 F-LOB, BrE). This allows for cross-time and cross-variety comparison with substantial reliability. The authors are deeply knowledgeable of the corpora, having been involved in compiling most of them, and tagging all of them for grammatical factors. Data from some additional corpora are included to provide greater time depth (Before LOB 1931 (B-LOB) allows comparison of changes in BrE to be extended back another thirty years) and comparison with spoken data (e.g. The Diachronic Corpus of Present-Day Spoken English (DCPSE), the spoken part of the British National Corpus (BNC), and the Longman Corpus of Spoken AmE (LCSAE)). As the authors aim to be meticulous in providing comparability, they are careful in their analyses to distinguish between evidence from the Brown family (including B-LOB), which are all constructed according to the same design, and evidence from other corpora, since these have not been constructed in similar ways.

The book consists of eleven chapters. The first two provide background information, eight provide detailed accounts of specific changes and the eleventh discusses what factors may have enabled the changes. There are three appendices, the third of which provides extensive additional statistical tables and charts.

Chapter 1 is entitled ‘Introduction: “grammar blindness” in the recent history of English?’ Here the authors address general issues concerning ‘the vast and complex system of rules which helps us organize words into constituents, clauses and sentences’ (p. 1) that linguists find, in contrast to the narrow, often variable collection of disputed usages that lay commentators hone in on. One of their interests in the book is to test evidence for recent and ongoing changes in English against current misperceptions, mainly among lay commentators, but also in some cases, among linguists, e.g. that the progressive is expanding with stative verbs. Another is to identify new developments that have not been noted before, within the domains they study. What precisely these are is not clear, but probably include the fact that core modals still outnumber semi-modals in both speech and writing (chapter 4).

Chapter 2 describes the corpora and lays out methodologies for achieving representativeness and comparability across time. Details of the annotation process used are provided, including use of automated tagging followed by manual post-editing. (Appendix II provides the tag-set used for the four corpora in the Brown family.) The shortcomings of the particular corpora are also discussed: the Brown family is small in size, texts are written (and edited) and somewhat restricted (e.g. no plays or advertisements are represented). Nevertheless their comparability makes them a unique resource. While corpus work is fallible and approximative, so, the authors point out, is ‘hypothetico-inductive/abductive’ work (p. 32). Arguments are made that quantitative and qualitative analyses need to be used in combination, since statistical findings regarding distributions require interpretation in terms of cognitive, social or historical factors.

The first study is the subjunctive mood (chapter 3). Subjunctive were is shown to be recessive in adverbial clauses such as counterfactual if-clauses, but less so in AmE than BrE (p. 67). This is the first example of what becomes a recurrent theme: differences, sometimes significant, between BrE and AmE either in frequency or in directions of change. In mandative expressions involving such verbs as require, demand, move (that), periphrastic should was used in over 80 percent of the examples in LOB but declined in F-LOB to 60 percent (p. 53). Use of the subjunctive, however, increased. While largely used with passives in BrE (Conditions have dictated that operations be scaled down (F-LOB, p. 59), there has been a statistically significant increase in the use of the subjunctive in active clauses. By contrast, subjunctive was already used in 85.9 percent of potential clauses in AmE by 1961; it shows only a small statistically insignificant increase to 90.3 percent in Frown (pp. 53–4).

Chapter 4 is devoted to the ‘core’ modals: will, would, can, could, may, might, shall, should and must, with ought (to) and need(n't) added as peripheral members. Needn't is contrasted with semi-modal need to, since it is followed by a bare infinitive and can occur with contracted negative (p. 72). It has long been observed that the core modals are in decline; it has also been hypothesized that they are being replaced by the semi-modals (be able/going/supposed to and be/(have) got/have/need/want to, (had) better), which are the subject of chapter 5. Chapter 4 shows that while the core modals are declining, the order of their frequency is the same in 1961 and 1991/2 (pp. 72–3), although there are slight differences in the rate of decline between BrE and AmE. Most striking is that four of the core modals, would, will, can and could have as a group lost only 4.7 percent of their usage in the thirty-year period, while those that were already less frequent by the end of the nineteenth century have declined far more rapidly: 22.7 percent as a group, with must, need(n't) and ought to each declining by over 31 percent (p. 73). The least common modals show diminishing functionality, and ‘paradigmatic atrophy’ (p. 80). For example, may is becoming primarily epistemic, and its permission meaning is being taken over by can; should is becoming almost exclusively modal (irrealis). However, must remains polysemous, with a slight increase in deontic uses (pp. 87–9). Epistemic uses of must remain robust and are not in serious competition with epistemic uses of have to, (have) got to or should (p. 116). Chapter 4 is one of the most interesting because of the insights it provides into obsolescence; it is also the densest, since the multiple polysemies of may, must and should are tracked.

The authors argue in chapter 5 that need to is not a variant of need(n't), as is often thought (e.g. Krug Reference Krug2000; Taeymans Reference Taeymans, Fischer, Norde and Perridon2004), but an independent development based on the main verb need and its related noun, as in a need to do something (p. 94). On this view, the development of need to is not a counterexample to grammaticalization as argued by Taeymans. Generally semi-modals increased, with be going to, have to and want to the forerunners (p. 99); however, be to has declined, especially in AmE. Be going to, have to and need to show some evidence of generalization of meaning, e.g. need to often invokes abstract and externally constrained, rather than inherent necessity (I'm not a feminist, but I do think you need to hear a balanced view of matters, F-LOB; contrast have to or must here, p. 110). The semi-modals are used significantly more frequently in speaking than in writing, and are more than four times as frequent in American speech than in the Brown family. All the same, core modals are still more frequent than the semi-modals by a ratio of 2 to 1 in spoken BrE (p. 79) and 1.8 to 1 in spoken AmE. The latter is, however, a considerable decrease from 7 to 1 in the 1961 Brown Corpus (p. 101).

Chapter 6 is devoted to the progressive, which has been increasing. The statistically most significant increase in both BrE and AmE is in the present tense, but only in the active voice in AmE. Another use increasing only in BrE is the progressive be-passive, mainly in the present tense. It is favored in genres with informational orientation, pertaining to current affairs, e.g. Press reportage (p. 137). There appears to have been no statistically significant increase with stative verbs such as hear, think, have, live, despite contrary claims by e.g. Aitchison (1991) (p. 130). Such an increase did, however, occur in the nineteenth century (Smitterberg Reference Smitterberg2005). Will + be -ing is identified as the only construction in BrE that ‘has significantly increased out of the whole range of future expressions surveyed’ (p. 139). Its use as volition-disclaimer (I'll be having my baby in June) is more frequent than use to refer to a situation that will be in progress at a reference time in the future (pp. 140–1).

Two types of passive, be- and get-passives, are discussed in chapter 7. Be-passives declined in both BrE and AmE formal written English (p. 153); however, within that decline, there is an increase in the proportion of by in BrE, and a decline in AmE. Unlike the be-passive, the get-passive has increased slightly; it is, however, a very infrequent alternative to the be-passive, in part, the authors hypothesize, because it has ‘adversative semantics’ (Did they know how wealth from over-large estates gets misused? LOB, p. 157). A third type of passive, the medio-passive (e.g. it wears well), is lexical and searchable mainly by individual lexical verbs (p. 160). Medio-passives appear to be spreading, but because they have generic meanings and are largely restricted to advertising copy, they are considered unlikely to compete straightforwardly with the be-passive (p. 164).

Expanded predicates are the topic of chapter 8. Given the wide range of verbs often included under this rubric, the authors limited their study to those with have, take and give + deverbal noun derived by conversion, e.g. take/have a run, but not have a thought. Two hypotheses put forward in previous studies are tested: that their frequency depends on text type, and that they have become more frequent over the last thirty years. The second hypothesis is not conclusively supported. There is ‘regional and variety-internal variation’ rather than ongoing change (p. 180). However, the first hypothesis is supported: the text type that most favors the expanded predicate is Fiction, especially narrative sections (p. 175). Prior studies had identified scientific writing and colloquial language as preferred text types. The authors attribute the differences in their findings to their narrow definition of the variable, which excludes expressions likely to occur in scientific writing (give consideration to something) (p. 175), and to restriction to the Brown family. On the other hand, spoken data outside the Brown family shows the predicates to be typical of informal, spoken English, especially BrE (p. 179).

A well-known long-term development in English has been the increase in non-finite clauses at the expense of finite ones (Los Reference Los2005), and, in the twentieth century of the gerund at the expense of the to-non-finite clause (e.g. Fanego Reference Fanego1996; Rudanko Reference Rudanko2006). Together, these changes have been called the Great Complement Shift (Rohdenburg Reference Rohdenburg, Dalton-Puffer, Kastovsky, Ritt and Schendl2006). Chapter 9 investigates spread of a subset of the complement shift, specifically of (i) start and stop in ‘catenative uses’ (start/stop to do/doing something), (ii) bare infinitives with help (helps (to) provide), (iii) from-less gerunds with prevent (prevent them (from) leaving) and (iv) want to in its transitional sense between ‘desire’ and modal ‘volition’ (That's one of the main problems with ‘God's Hands’ . . . You want to like these people. . . . But the whole thing feels like an exercise in self-indulgence, Frown, p. 201). The change in (i) occurred primarily between 1931 and Brown 1961, the other changes between LOB and F-LOB (p. 201). There is a slight (but statistically non-significant) increase in to-infinitives overall, but mainly in Press texts (pp. 202–3). The increase in bare infinitives with help shows convergence between BrE and AmE. However, absence of from after prevent has become a distinct Briticism (p. 193). The authors point to a large discrepancy between the small numbers of modal uses of want to in chapter 4 and the ‘massive short-term increase’ in want to overall (p. 200). Most uses are ambiguous semantically between the main verb and the auxiliary (a distinction rarely made; see e.g. Krug Reference Krug2000). ‘Short-term’ presumably refers to the authors’ expectation that the semi-modal use will crystallize after the 1991 ‘“hot” phase of dynamic change’ (p. 201) on the assumption that increase in frequency leads to grammaticalization (Bybee Reference Bybee, Joseph and Janda2003). This is an ongoing change worth following in the next few decades to find empirical evidence for hypotheses about the relationship between frequency and grammaticalization.

The final data chapter, chapter 10, concerns aspects of the noun phrase. The main conclusion is that the noun phrase, unlike the auxiliary, is undergoing ‘densification’ defined as ‘an increase in the proportion of content words to function words’ (p. 208). This is manifest in the decline of N1 of N2 phrases (the fruit of the coconut palm) compared to N2's N (the coconut palm's fruit) and to N2 N1 (coconut palm fruit). This change illustrates loss of both preposition and article (p. 209). Densification of nouns had already occurred in Press texts by the mid 1960s, and had reached a ‘noun saturation point’, interpreted as a point at which information density, leading to economy, ceased to be viable in its competition with ‘popularization’ and the need to maintain a readership (Biber Reference Biber, Aitcheson and Lewis2003). The increases in the last thirty years have been primarily in General Prose (p. 215). Increases in subtypes of nominals are most notable in use of acronyms and proper nouns and noun sequences with plural attributive nouns (fisheries protection) (p. 221).

A second aspect of the noun phrase discussed in chapter 10 is the choice of relativizer (wh-pronouns, that or zero). This does not affect densification, but interfaces with informational specificity. Given the multiple uses of that, wh-relatives ‘can bear a heavier functional load’, and can be used with the greatest variety of syntactic functions (p. 227). What appears at first to be a decline in wh-relatives in AmE is in fact due to the loss of which (p. 229). Over the thirty years there has been a decline in pied-piping and a rise of stranding in BrE. In AmE stranding with zero relatives has risen 50 percent (things we could perfectly well do without, LOB, p. 233). Data of this kind have been of central interest to researchers on parsing (e.g. Wasow Reference Wasow2002; Hawkins Reference Hawkins2004).

Chapter 11 summarizes hypotheses about functional explanations for the changes (misleadingly called ‘determinants’ – they are really enabling factors) that have been mentioned briefly throughout the book. One is grammaticalization. All of the changes under discussion began before the 1960s, so what is at stake here is evidence of the way in which changes become entrenched in terms of token frequency, weakening/generalization of meaning (e.g. be going to) and coalescence (increased use of wanna, gotta). The changes attest to the importance of frequency not so much at early stages of a grammaticalization but when the new item is ‘being integrated and spread through the system’ (p. 269) or when one item becomes a ‘magnet of attraction’ (Krug Reference Krug2000: 243). Also at stake is the decline of grammaticalized expressions (core modals) (pp. 238–9).

Sometimes increased grammaticalization is linked to text category or style. Particular attention is paid to colloquialization, the shift toward a more speech-like style (Mair Reference Mair, Hickey and Puppel1997). The trend toward ‘spontaneous directness’ in writing (p. 239) is correlated with the rise in General Prose and especially Learned writing of some semi-modals, the progressive, contracted negatives and modals, and decline of wh-relatives. However, there is also anti-colloquialization (p. 245) in densification of the noun phrase. Drawing on Biber (Reference Biber, Aitcheson and Lewis2003), the authors point out that there are competing demands. A tendency to economize by packing more information into less space competes with ‘selling’ it in a more informal style (p. 252). A further factor is ‘democratization’. While democratization is connected with decreased formality (colloquialization), the authors distinguish it as involving ‘reduction of overt markers of power asymmetry’ (Fairclough Reference Fairclough1992: 98) (p. 259). It affects the noun phrase with respect to naming and (non)use of titles such as President, Mr or even surnames.

Another trend discussed is Americanization. In most cases AmE is ahead of BrE, e.g. in the declining frequency of core modals, be-passives and wh-relatives, or in the increasing frequency of epistemic have to, help to and the -s genitive. But sometimes BrE is ahead, e.g. in the increase in semi-modals as a whole and progressive passive (p. 253). The authors point out that such frequencies do not necessarily imply transatlantic influence, only (partially) similar developments. An exception is the mandative subjunctive, which declined in earlier BrE but then increased, converging toward AmE in F-LOB. While Americanization appears to lead with colloquialization in many respects, there is a confound: the tendency for prescriptivism in America which may have promoted decrease in which-relativization and the be-passive.

Finally, the authors argue that the drift toward analyticity that Sapir (Reference Sapir1921: 156–70) projected for English is not strongly supported by the Brown family. Whom is showing few signs of decline, and the s-genitive has been expanding (p. 265). They suggest that drift should be rethought as ‘frequency drift’, and ways in which it feeds into making peripheral combinations less peripheral and central ones more central (p. 270). The chapter ends with a summary table showing correlations between the explanatory factors and features that increased or declined in frequency over the last thirty years of the twentieth century, as evidenced by the Brown family of corpora.

CCE is exceptionally rich in detail and thoughtful proposals about why certain changes may have taken place. It is the outcome of several decades of work on the part of all the authors, especially Geoffrey Leech. Christian Mair's book on Twentieth-Century English (Reference Mair2006) discusses changes in lexicon and pronunciation and also briefly previews most of the grammatical changes central to CCE as well as colloquialization and Americanization. Earlier studies contributing to this one have explored a range of different corpora and other parts of the English-speaking world (e.g. Hundt Reference Hundt1998 on New Zealand English). Beyond the many examples from the corpora, what is new is chiefly the rigor of the comparisons, the development and use of grammatical tagging, and the extension of LOB back to the 1930s.

While acknowledging the limitations of the Brown family, the authors convincingly show that it allows for a uniquely detailed and comprehensive comparison of different stages. The Brown family constitute a ‘small’ corpus since each subcorpus is roughly a million words, and together they total about 4 million words. They are greatly overshadowed in size by current large corpora such as COHA (400 million words). Recent exponential increases in the size of computerized corpora have raised questions about the advantages and disadvantages of different sized corpora. ‘Small’ corpora are clearly most valuable for grammatical changes of the sort discussed here, where the numbers are large and significance tests can be conducted. Indeed, they are probably the only reliable source of information for such changes, since only a selected subset of examples could be analyzed in detail from a corpus such as COHA. Furthermore, it is impossible to do more than automatic grammatical tagging of such large corpora: the fine-tuning with manual post-editing is not feasible, and therefore the analyses become less precise. The large corpora, however, are ideal for the study of changes with lower frequencies, such as the Great Complementation Shift. Given sufficiently rigorous criteria for selection of data (and funding), they can also be extended as ‘monitor corpora’ tracking new developments (Davies forthcoming).

CCE suggests a number of issues that will no doubt inspire much research in the future, not only in English, but in any language for which electronic corpora are available over a fifty- to hundred-year period. I can mention only a few here.

One has to do with obsolescence. While much work on grammaticalization has focused on increases in frequency (Bybee Reference Bybee, Joseph and Janda2003) and expansion of syntactic, host-class and semantic-pragmatic factors (Himmelmann Reference Himmelmann, Bisang, Himmelmann and Wiemer2004), it has also been a basic but largely unexplored assumption that grammatical items disappear either because of phonological reduction or, more frequently, because they are no longer used, and that they are renewed. CCE provides solid empirical evidence that competition accompanies decline, rather than following it (see Lehmann Reference Lehmann1985). As the authors say, ‘there needs to be some theory of linguistic decline explaining a long-drawn-out process of losing frequency’ (p. 238). This theory needs also to account for the different ways in which competition plays out. With a large-size corpus like COHA one might, for example, be able to study the decline of that-complementation, compare it to that of the modals, and ask whether there is comparable evidence of restriction to certain functions.

Another question is how to understand the notion of what it means to reach ‘saturation point’ in a model of change as usage change. This issue comes up in connection with the fact that neither the subjunctive (p. 54) nor densification of the noun phrase in the Press genre (p. 210) have undergone much increase in AmE, while both factors have increased substantially elsewhere (the subjunctive in BrE, densification in Learned and General Prose genres). There is often discussion in sociolinguistics and in grammaticalization of changes ‘going to completion’, but this is a problematic concept, because, barring ‘obligatorification’ (Lehmann Reference Lehmann1955) or complete loss, it is not clear what it might mean. The authors suggest that categorical, i.e. 100 percent, use is not to be expected when they say ‘the expectation that the subjunctive would become obligatory is probably an unreasonable one’ (p. 54), but do not say why. Implications for theories of categoriality, gradience and change (see Traugott & Trousdale Reference Traugott and Trousdale2010) deserve to be followed up.

The quantificational data presented in CCE is supplemented with a rich set of hypotheses about large-scale factors that might have led to changes, from grammaticalization to democratization. With the development of fine-grained tools such as regression analyses or diachronic collostructional analysis (e.g. Hilpert Reference Hilpert2008), it should be possible to investigate more fully what smaller-level factors interact with those under review in domains less well studied than the polysemies of the (semi-) modals. For example, Rosenbach (Reference Rosenbach2010) shows that the second and third modifier constructions in the change that the authors designate in chapter 10 as a shift from (i) N1 of N2 phrase to (ii) N2's N1 and to (iii) N2 N1 have very different functions. (ii) is a determiner construction (the FBI's director), (iii) a classifier construction. In the ARCHER corpus there has been increase in collective (FBI director) and animate classifiers (Obama supporters) since the mid nineteenth century (Rosenbach Reference Rosenbach2010: 168). This suggests that functional changes have interfaced with densification.

One of the authors’ objectives was to test the assumption that ‘the spoken language is the leading force in bringing about change in the standard language’ (p. 45). Since none of the changes discussed originate within the period under discussion, no evidence about origination emerges. However, spoken corpora supplementary to the Brown family show that many of the changes under discussion (both increase and decline) have been occurring more rapidly in the spoken than the written language. This is particularly true of the increase in semi-modals and of the progressive. A few changes are, however, mainly found in writing, e.g. the rise of the subjunctive. While quantitative analysis may show the numbers, many more corpora of spoken language will need to be developed and newly emerging constructions will have to be identified before it will be possible to fully test the hypothesis that spoken language is ‘the leading force’ (where both ‘the’ and ‘force’ are taken at full value). As the authors point out, there are many confounds. Counterfactuals (where some of the subjunctives appear in competition with would) are simply not the stuff of speech; describing situations as they unfold (the realm of the progressive) is.

As more corpora are developed it will be possible to see how and why some radical changes occur in the wake of external changes such as the rise of internet usage. The authors see the stability of the ‘genre map’ over the thirty years under discussion as a distinct advantage in achieving comparability across time (p. 30), but also acknowledge that the limitation of CCE to standard UK and US English at a time of globalization (mentioned briefly p. 259) is a drawback. While the genre map did not change much in the twentieth century, it has clearly changed enormously in the last decade since Frown and F-LOB. It did so in Early ME and the Renaissance as well. How can rigorous quantitative corpus work be designed for periods of such instability in the ‘genre map’, and how will it affect any comparable study made in 2021?

Regardless of any limitations of the corpora, the authors have developed a rigorous methodology for tagging, quantifying and analyzing electronic corpus materials, and revealing the multifactorial nature of change in use. There has been a shift in the last decade to studying change as usage change (see Croft Reference Croft2000; Bybee Reference Bybee2006). The authors have shown what kinds of empirical evidence are required for this approach. CCE will be foundational for future work on language change in progress, not only expansion, but also decline, and how competition within constructional space progresses.

CCE is well edited. The index is excellent as a subject index. It lists a few names, but a full index of names would have been helpful. Lack of a corpus bibliography is surprising, but references can be found in the Preface and chapter 1.