Table 10 shows the frequencies of the semi-modals which were the subject of inves- tigation in Leech et al. (2009: 91–117).18 In their follow-up work to this study using
17. The figures given for Kolhapur are the raw frequencies in the whole corpus. The ICE figures are based on the 200 written text samples and have been normalised for com- parison. (The raw figures – 312 for must and 127 for shall – were used to test for statistical significance.)
A further complicating factor in comparisons of ICE and the Brown family is the difference in the type and amount of the written genres sampled, particularly at the informal end of the spectrum. For instance, a Brown family corpus contains more fictional texts than an ICE, whereas only ICE corpora sample unpublished writing such as student essays and personal letters.
18. Consisting of more than one element, the semi-modals are more difficult to search for than the modals, and issues of precision and recall arise in the corpus searches. For example, the figures for be able to in Table 10 give the results of a (manually post-edited) search for the sequence able to, plus another search for able [any word] to. Examples in which able … to was separated by more than one word would have been missed. A particular problem was posed by the interrogative forms of be to. Here, a search for [form of be] [any word] [any
www.ebook777.com
Free ebooks ==> www.ebook777.com
136 Christian Mair
B-LOB, Leech and Smith (2009: 187–192) and Leech (2011, 2013) deal with the semi- modals only selectively. The present study is therefore the first to provide coverage of the semi-modals in B-LOB and B-Brown at the levels of Leech et al. (2009: 91–117).
Tables 10 and 11 do not give totals, because – unlike the modals, which are a closed category characterised by clearly defined structural properties – the semi-modals form a structurally far more heterogeneous and potentially also more open-ended set of constructions:19
Table 10. Frequency of semi-modals in B-LOB, LOB and F-LOB B-LOB
(1930s) LOB
(1961) Change %
B-LOB→LOB F-LOB
(1991) Change %
LOB→F-LOB Change % overall
be able to 247 246 –0.4% 248 +0.8% +0.4%
be going to 205 248 +*21.0% 245 –1.2% *+19.5%
be supposed to 33 22 –33.3% 47 **+113.6% +42.4%
be to 494 451 –8.7% 376 *–16.6% ***–23.9%
(had) better 23 50 **+117.3% 37 –26.0% +60.9%
(have) got to 31 41 +32.3% 27 –34.1% –12.9%
have to 505 757 ****+49.9% 825 +9.0% ****+63.4%
need to 20 53 ****+165.0% 194 ****+266.0% ****+870.0%
want to 259 357 ****+37.8% 423 **+18.5% ****+63.3%
(LOB and F-LOB figures from Leech et al. 2009: 286; significances [log likelihood]:
* = p < 0.05, ** = p < 0.01, *** p < = 0.001, **** p < 0.0001)
word] to would not only have caught is he to be commended/is someone to be commended/
is the councillor to be commended etc. (= high recall), but also a vast amount of false hits of the type is necessary to/is here to/is in London to/is always eager to etc. (= low precision).
The compromise adopted was to search for [am/are/is/was/were] [I/you/he/she/it/we/you/
they] to, which yielded 25 relevant examples from B-LOB and 9 from B-Brown. Examples in which a word other than a personal pronoun was the subject or in which the subject consisted of two or more words would have been missed. The same strategy was employed to look for have-to interrogatives with auxiliary syntax (have we to obey?). It yielded no unambiguous examples fit for inclusion in the counts for either corpus. For a borderline case, see (1) below.
19. Note, for example, that Leech et al. include be able to as a possible alternative to can, but do not take into account be allowed to, which is in variation with both can and may in their
‘permission’ uses. An item which could have been included alongside be going to and be to is be about to.
Free ebooks ==> www.ebook777.comCross-variety diachronic drifts and ephemeral regional contrasts 137 Table 11 presents the distribution of the same semi-modals in the American corpora.
Table 11. Frequency of semi-modals in B-Brown, Brown and Frown B-Brown
(1930s) Brown
(1961) Change %
B-Brown→Brown Frown
(1991) Change %
Brown→Frown Change % overall
be able to 189 191 +1.1% 202 +5.8% +6.9%
be going to 170 216 *+27.1% 332 ****+53.7% ****+95.3%
be supposed
to 46 48 +4.3% 55 +14.6% +19.6%
be to 425 344 **–19.1% 217 ****–36.9% ****–48.9%
(had) better 20 41 **+105.0% 34 –17.1% +70.0%
(have) got
to 36 45 +25.0% 52 +15.6% +44.4%
have to 362 627 ****+73.2% 639 ****+1.9% ****+76.5%
need to 35 69 ***+97.1% 154 ****+123.2% ****+340.0%
want to 277 323 +16.6% 552 ****+70.9% ****+99.3%
(Brown and Frown figures from Leech et al. 2009: 286; significances [log likelihood]:
* = p < 0.05, ** = p < 0.01, *** p < = 0.001, **** p < 0.0001)
Among the high-frequency items, be able to remains largely stable in BrE and AmE. Have to and be going to show the expected overall rise in both varieties, although there is a difficult-to-account-for damper in the most recent British data for going to.
As regards interrogative uses of have to, subject-auxiliary inversion – recorded as a
“somewhat old-fashioned” British option in Quirk et al. (1985: 145) – is attested just once even in B-LOB, the oldest British corpus:
(1) What on earth had she to be afraid of? (B-LOB, K)
Note that syntactic extraction contexts of this type (wh-questions) are potentially ambiguous between modal-obligation (she had to do something) and possessive- existential readings (she had something to do). The present example, and two more found in the corpora, are contextually plausible only in the latter, non-modal interpretation.
In both sets of corpora, the highest log-likelihood values are obtained for need to (i.e. need with main-verb syntax in modal function) – a form whose dramatic spread in the very recent history of English has been noted in numerous studies (Taeymans 2004; Nokkonen 2006; Müller 2008; Leech & Smith 2009: 191). Such drastic increases in a very short period reflect the dynamic phase of ongoing grammaticalisation, in which semantic bleaching, the increasing conventionalisation of new grammatical patterns and their spread through all registers and text-types conspire to boost cor- pus frequencies. By comparison, going to and have to represent mature instances of
www.ebook777.com
Free ebooks ==> www.ebook777.com
138 Christian Mair
grammaticalisation, but here too the figures show that these grammatical innovations of past centuries are still spreading forcefully in written English in the present.
The most striking contrary development in both varieties is the rapid obsoles- cence of the semi-modal be to.20 As Hundt (2013) has shown, this construction has already shed a formerly current non-finite form (cf. a ship being to sail the next day:
see Hundt 2013: 7 for the full example). In present-day English it is still productive in a wide range of usually rather specialised uses relating to the expression of modal- ity and future time (Declerck 2010). Neither the gains nor the losses documented in the extended Brown family are likely to be restricted to BrE and AmE. The question of how long a structurally and functionally complex semi-modal such as be to can maintain its full grammatical productivity in the face of statistical attrition is thus one which would be worth asking for any New English, as well.
Interesting findings are obtained for have got to. This form, though fully gram- maticalised, has never become as common as the related have to. It is rare in writing in all varieties. In speech, it is somewhat more common in BrE than in most other variet- ies, including the British-derived New Englishes. Consider, for example, the following figures from the private conversations in ICE-GB, the British component of the Inter- national Corpus of English, in comparison with CSAE, the “Santa Barbara” Corpus of Spoken American English (Seggewiò 2012: 199).
Table 12. Must, have to, have got to, need to21 in ICE-GB (private conversations) and CSAE (Seggewiò 2012: 199)
ICE-GB abs. ICE-GB pmw CSAE abs. CSAE pmw
must 52 253 24 96
HAVE to 140 681 203 815
(HAVE) got to 87 423 89 357
NEED to 65 316 89 357
In interpreting her figures, Seggewiò (2012: 128–130, 199) shows that if the analysis is restricted to affirmative contexts 62 (= 302 pmw) examples of have got to remain in ICE-GB, but only 41 (=165 pmw) in CSAE. More importantly, however, the CSAE examples display advanced stages of reduction, with almost 90 per cent show- ing phonetic contraction of got to → gotta, and more than 70 per cent lacking a form of the auxiliary have. Owing to different transcription conventions, contraction rates
20. Nesselhauf 2006 asks whether the decline of be to correlates with the rise of going to.
21. Need with auxiliary syntax does not occur in either corpus.
Free ebooks ==> www.ebook777.comCross-variety diachronic drifts and ephemeral regional contrasts 139 are difficult to obtain in ICE-GB, but in striking contrast to the spoken data from AmE have remains present in slightly over 90 per cent of the British examples. As Seggewiò puts it: “Paradoxically […], though being further grammaticalised in AmE, have (got) to occurs less frequently” (2012: 128). This observation ties in well with work by Jankowski (2004) and Tagliamonte & D’Arcy (2007), who have shown that (have) got to has been on the decline in North AmE from the mid-20th century. None of these developments in spoken English are reflected in the Brown family statistics, where we observe a statistically insignificant decline of have got to in BrE and an equally insignificant increase in AmE. The frequencies are modest throughout, so that – if anything – the data show that (have) got to is a marginal presence in written English both in Britain and the United States. Although statistically significant patterning is absent in the data, some individual examples are still interesting. Many of the 1930s US examples from B-Brown involve represented speech from works of fiction (see (2) below), and one (example (3)) even has auxiliary do in the tag referring back to got:
(2) Nell said, “You got to work ‘em. You got to keep them guessing, but don’t give in to ‘em unless they make you.” (B-Brown, K15) (3) When we hit a mite of a grade I walk. If the grade gets steep I push. I got to
have a horse, don’t I? (B-Brown, N2)
A tendency to use do-support with (have) got (to) – in other words a re-analysis of a semi-modal construction as a finite verb – can be documented for informal and non- standard North AmE from the late 19th century (Mair 2012). Examples such as (3) illustrate a bridging context which helps us understand how auxiliary do can be gradu- ally adapted for this new use.