An obvious starting point for an analysis of modality in the extended Brown family is to look for regional contrasts between BrE and AmE at three successive diachronic sampling points (c. 1930, 1961, 1991/92). All the modals and semi-modals investigated here are attested in BrE and AmE, but nevertheless the relevant literature suggests that there are contrasts in the way they are used on either side of the Atlantic. For example, Trudgill and Hannah claim that “several of the modals are used with a different fre- quency or meaning in N[orth]Am[erican]Eng[lish] than in Eng[lish]Eng[lish]” (2008 [1982]: 62). Among the more robust tendencies they record (2008: 62–63) is the per- sistence of shall and shan’t as future markers in the first person in more conservative and formal BrE (I shall tell you later, I shan’t be able to come), but also the use of shall in questions of the type Shall I drink this now? (where AmE is claimed to prefer should).
Similarly, should is said to continue as an alternative to would in the first person of the
5. Endonormative stabilisation is considered Stage 4 in Schneider’s (2007) five-stage model of the nativisation of (post-)colonial Englishes.
Free ebooks ==> www.ebook777.comCross-variety diachronic drifts and ephemeral regional contrasts 125 conditional in BrE, but not in AmE (I should enjoy living here if I could afford to do so).
The use of need with full auxiliary syntax is also claimed to be far more common in BrE than in AmE. On the other hand, would as indicator of past habitual aspect (When I was young, I would go there every day) is said to be dominantly American.
Table 1 shows how BrE and AmE compare in the 1930s with regard to the use of modal verbs.6 When interpreting the figures, it is necessary to bear in mind that the regionally specific uses mentioned by Trudgill and Hannah account for a large part of all examples in the case of low-frequency modals such as shall and need, but that this is not so for medium- and high-frequency ones such as should and would. Past habitual would, for example, is so rare in comparison to the more common uses of this modal verb that it will have a negligible impact on the totals. Here and in most following tables, levels of statistical significance are represented schematically through an asterisk-based notation. The method of calculation and absolute p-values are docu- mented in the Appendix.
Table 1. Nine core modals and two marginal modals in B-LOB and B-Brown7
B-LOB B-Brown Log likelihood
would 2,673 2,412 ***10.30
will 3,055 2,606 ****30.19
can 2,039 1,718 ****23.54
could 1,433 1,332 2.52
may 1,702 1,357 ****34.74
should 1,486 1,037 ****74.74
must 1,265 955 ****39.59
might 713 626 *4.61
shall 475 289 ****43.41
ought (to) 135 111 2.05
need(n’t) 94 49 ***13.84
Total 15,070 12,492 ****209.92
(significances: * = p < 0.05, ** = p < 0.01, *** p < = 0.001, **** p < 0.0001)7
6. The counts in this table, as in all the following ones, cover the relevant contracted forms and enclitic negations. Thus, the figure for would includes instances of ’d and wouldn’t.
Instances of ’ll have been added to will rather than shall.
7. While Leech and Smith distinguished three levels of significance (* = p < 0.05, ** = p < 0.01,
*** p < = 0.001), the present study adds a fourth, **** p < 0.0001, to highlight the statistically most robust trends.
www.ebook777.com
Free ebooks ==> www.ebook777.com
126 Christian Mair
At first sight, these figures are suggestive of pervasive and significant regional con- trasts. However, by the 1960s (Table 2, figures from Leech et al. 2009: 74, 283), much of this variability seems to have disappeared:
Table 2. Nine core modals and two marginal modals in LOB and Brown
LOB Brown Log likelihood
would 3,032 3,053 0.0
will 2,822 2,702 3.46
can 2,147 2,193 0.23
could 1,741 1,776 0.16
may 1,333 1,298 0.73
should 1,301 910 ****72.12
must 1,147 1,018 **8.57
might 779 665 **9.78
shall 355 267 ***13.08
ought (to) 103 69 **6.99
need(n’t) 76 40 ***11.60
Total 14,836 13,991 ****30.68
(significances: * = p < 0.05, ** = p < 0.01, *** p < = 0.001, **** p < 0.0001)
What has remained is the over-representation of shall, should and full-auxiliary need in BrE. This over-representation is also evident in the synchronic comparison of the 1990s corpora F-LOB and Frown (Table 3, figures from Leech et al. 2009: 74, 283).
In addition, it is interesting to note that some of the 1930s regional contrasts seem to
“resurface”, for example with will and may.
The question raised by the findings assembled in Tables 1 to 3 is whether there really was a robust regional contrast between BrE and AmE in the 1930s which was temporarily levelled by the 1960s and then re-asserted itself at more modest levels in the 1990s. Such a scenario would be plausible only if we were able to argue that the regional origin of a writer in Britain or the US was the major causal determinant of the variation observed in the Brown family corpora. However, this is not so. As will be shown, cross-variety diachronic drift and genre are confounding factors whose influ- ence is much stronger than region. A considerable portion of the English modal sys- tem (modal verbs and semi-modals included) is involved in a diachronic drift which affects most major varieties of the language. In such a situation, it is very unlikely that stable regional differences will emerge and be maintained. Rather, regional con- trasts at any given time will be temporary and ephemeral, because they result from the respective varieties moving in the same direction and towards the same ultimate goal at slightly different speeds.
Free ebooks ==> www.ebook777.comCross-variety diachronic drifts and ephemeral regional contrasts 127 Table 3. Nine core modals and two marginal modals in F-LOB and Frown
F-LOB Frown Log likelihood
would 2,682 2,868 *5.84
will 2,708 2,402 ****19.0
can 2,213 2,160 0.76
could 1,767 1,655 *3.91
may 1,100 878 ****25.45
should 1,148 787 ****68.53
must 814 668 ***14.72
might 640 635 0.03
shall 200 150 **7.28
ought (to) 58 49 0.78
need(n’t) 44 35 1.05
Total 13,374 12,287 ****48.45
(significances: * = p < 0.05, ** = p < 0.01, *** p < = 0.001, **** p < 0.0001)
This is not expected to be different in the New Englishes. As a demonstration case, consider Table 4, which lists the frequencies for must and shall in the two Antipodean Brown clones (ACE, the Australian Corpus of English, and WCWNZE, the Wellington Corpus of Written New Zealand English), comparing them to F-LOB and Frown, the diachronically closest constituent corpora of the core Brown family.8
Table 4. Must and shall in ACE, WCWNZE, F-LOB and Frown9
ACE9 WCWNZE F-LOB Frown
must 636 685 814 668
shall 113 132 200 150
(significances: ACE vs WCWNZE not significant; ACE vs F-LOB p < 0.0001 = ****; ACE vs Frown p < 0.001 = * [shall only]; WCNZE vs. F-LOB p < 0.01 = ** [must], p < 0.001 [shall]; WCNZE vs Frown not significant)
8. The sampling year for ACE was 1986. WCWNZE comprises texts published between 1986 and 1990.
9. ACE figures were obtained from ACE II, which does not contain all the 500 two- thousand- word samples of the complete corpus, but only the 375 for which the compilers were able to obtain copyright clearance. ACE II contained 477 instances of must and 85 of shall. The fre- quencies in Table 4 are normalised (n per million words). The absolute frequencies were used as input for the significance tests.
www.ebook777.com
Free ebooks ==> www.ebook777.com
128 Christian Mair
As a synchronic “snapshot,” Table 4 allows the following observations: (i) there are no significant regional contrasts between written AusE and New Zealand English (NZE), (ii) both Antipodean varieties differ from AmE in minor ways only, and (iii) the only pronounced contemporary regional contrast is between BrE and the other three varieties. One possible explanation for these observations is “Americanisa- tion”: the global pull of American norms, it could be argued, has prised historically British New Englishes out of their orbit and led to a re-orientation of usage norms towards those of the group which currently dominates the English-speaking world demographically, economically and culturally. Given the demonstrated volatility of the regional contrasts in the Brown family, however, this explanation rests on shaky his- torical foundations. Diachronic studies are needed for AusE and NZE – firstly, in order to confirm that the absence of noteworthy regional contrasts between AusE and NZE holds for older historical sampling points and, secondly, to find out whether the pres- ent similarities between AusE, NZE and AmE have come about gradually, as the result of a long-term historical drift set in motion long ago and operative in all varieties, or recently and suddenly, through dialect contact and exposure to North AmE. A recent study by Collins (2014), covering selected modals and semi-modals in AusE fiction from 1800 to 1999, is a valuable first step in this direction (see also Hundt this volume;
Collins this volume; Yao this volume).
Without a diachronic reference point, it is also difficult to interpret the frequen- cies for must and shall found in the Kolhapur corpus of Indian English, which contains texts sampled in the year 1978. This corpus contains 766 instances of must and 364 of shall. This is a solid British-type statistical profile (coming close to F-LOB for must and to LOB for shall), so that the issue of Americanisation does not arise. On the other hand, a comparison with more recent data from ICE India (see Section 3 below) will show that developments in Indian English partly moved away from BrE, following a dynamic of their own.
3. Some core modals are on the decline, but the category remains robust: