OUP UNCORRECTED PROOF – REVISES, 06/14/12, NEWGEN
Chapter 15
CHANGE IN
THE ENGLISH
INFINITIVAL PERFECT
CONSTRUCTION
Jill Bowie and Bas Aarts
1. Introduction
The availability of searchable electronic corpora composed of textual material from
different time periods has made studying change in the English language easier.
However, there are a number of methodological dimensions to the use of corpora
in the study of current change. For some “big is better” (Davies, this volume), while
for others “small is beautiful” (Hundt and Leech, this volume; Smith and Leech,
forthcoming). While acknowledging the distinct advantages of using large corpora, we believe that detailed analysis of small corpora, especially those that are
parsed, can reveal trends that may be missed by other approaches. Furthermore,
spoken language corpora are particularly valuable for studying short-term change:
spoken language is primary, and grammatical changes are likely to manifest themselves in that medium first.1
1 We are very grateful to Sean Wallis for assistance with the statistical analysis in this
paper. We also acknowledge with gratitude the support of the Arts and Humanities
Research Council under grant AH/E006299/1.
16_Nevalainen_Ch15.indd 200
6/14/2012 2:37:29 AM
OUP UNCORRECTED PROOF – REVISES, 06/14/12, NEWGEN
change in the english infinitival perfect construction
201
The study of recent grammatical change presented here draws on the Diachronic
Corpus of Present-Day Spoken English (DCPSE), based at the Survey of English
Usage, UCL. DCPSE is unique in several ways. First, it contains exclusively spoken
(and mainly spontaneous) English, in two subcorpora with matching text categories, allowing diachronic comparison over a 30-year span. The earlier subcorpus
contains approximately 464,000 words from the London–Lund Corpus (LLC) from
the late 1950s to 1970s, while the later subcorpus contains around 421,000 words
from the British Component of the International Corpus of English (ICE-GB) collected in the early 1990s. Second, the corpus is fully parsed and searchable with
dedicated corpus exploration software called ICECUP (International Corpus of
English Corpus Utility Program).2
Our study focuses on the British English (BrE) perfect construction, and in
particular the infinitival perfect. We first compare the inflectional subtypes of the
perfect (present/past/nonfinite) in terms of changing frequencies of use, before
examining the infinitival perfect in more detail. There are several reasons for investigating the perfect. The first is to extend what is known about its longer term historical development. From early origins in Old English, the perfect increased markedly
in frequency through Middle English into Early Modern English, but this advance
later appears to have been halted and even, after 1800, reversed, at least in American
English (AmE) (Elsness 1997; Fischer and van der Wurff 2006). Second, contemporary written sources show that the perfect is less frequent in AmE than in BrE
(Elsness 1997; Hundt and Smith 2009);3 it is therefore worth investigating whether
British usage is changing toward American norms as appears to have happened
with other grammatical features (Leech et al. 2009), since the spoken language may
be “ahead” in this regard. Finally, the perfect provides a case study of a construction
whose subtypes show distinctive patterns in terms of relative frequency, semantic
specialization, and syntactic contexts of occurrence. Such factors are likely to influence changing patterns of usage, making a fine-grained analysis valuable.
2. The perfect construction
in DCPSE
The English perfect construction involves the perfect auxiliary have followed by
a verb in the past participle form. It occurs in present, past, and nonfinite forms,
all of which typically function to express anteriority (i.e. pastness relative to a
reference point). The present perfect (I have read the book) generally presents a
2 See Svartvik (1990) on LLC and Nelson, Wallis, and Aarts (2002) on ICE-GB/ICECUP;
see also http://www.ucl.ac.uk/english-usage.
3 Hundt and Smith (2009) look at the present perfect; Elsness (1997) also looks at other
perfect subtypes.
16_Nevalainen_Ch15.indd 201
6/14/2012 2:37:32 AM
OUP UNCORRECTED PROOF – REVISES, 06/14/12, NEWGEN
202
rethinking evidence
Figure 1. FTF for perfect infinitive auxiliaries
situation as occurring within (or even continuing through) a time span beginning
in the past and leading up to the present, with an added dimension of “current relevance” (i.e. a focus on the present repercussions of the situation). The past perfect
(By the time he returned I had read the book) typically encodes anteriority to a past
reference point, and the two nonfinite subtypes (infinitival and -ing-participial)
indicate anteriority in various types of construction (He must have read the book;
Having said that, I still like her). The present perfect has developed a specialized
use through contrast with the (morphologically marked) past tense (used to present a time as wholly in the past rather than connected to the present). This contrast, however, is neutralized in the other perfect subtypes, which can correspond
to either a present perfect or a simple past (Huddleston and Pullum 2002: 146).
Examples of the perfect construction can be retrieved in DCPSE by using
Fuzzy Tree Fragments (FTFs), a search facility within ICECUP that allows users to
construct partial tree diagrams and to choose the level of detail specified (Aarts,
Nelson, and Wallis 1998; Nelson, Wallis, and Aarts 2002). Figure 1 shows a simple
FTF search for a single node of category “auxiliary” with type feature “perfect” and
inflectional feature “infinitive”. Categorial information (e.g. word or phrase class)
is displayed in the upper right segment, functional information (such as subject,
NP head) in the top left segment, and additional features in the lower segment. In
this instance the function has been left unspecified.
The combination have+got requires special attention. “Semi-modal” have
got [to] (as in a lot of work has got to be done) is unproblematic: it is analyzed in the
corpus as an auxiliary with type feature “semi” and so automatically excluded by
FTF searches for perfect auxiliaries. However, the searches do include, alongside
clear instances of the perfect construction with got (e.g. How advanced have they
got?), instances where the combination takes an NP object and expresses a stative
meaning, such as he’s got two kids (‘he has two kids’). These represent an idiom
historically derived from a perfect construction but now semantically distinct
(Huddleston and Pullum 2002: 111–13). Although they are sometimes included in
corpus counts of the perfect (e.g. Biber et al. 1999: 463–67), they are best excluded.
We therefore used FTFs to find instances where have, parsed as a perfect auxiliary, is followed by got (allowing for intervening material such as adverbs). These
16_Nevalainen_Ch15.indd 202
6/14/2012 2:37:32 AM
OUP UNCORRECTED PROOF – REVISES, 06/14/12, NEWGEN
change in the english infinitival perfect construction
203
Table 1. Frequencies of perfect auxiliaries in DCPSE, divided by inflectional
category
Inflectional
category
LLC
ICE-GB
pmw
%
A: χ2
(words)
B: χ2
(perfect)
2,523
6,241.22
0.75%
0.07
18.82 s
1,767.04
473
1,170.07
−33.78%
50.92 s
26.84 s
652
1,463.93
416
1,029.07 −29.70%
31.88 s
14.76 s
78
175.13
58
143.48
−18.08%
1.33
0.25
4,276
9,600.86
3,470
8,583.84
−10.59%
24.04 s
pmw
raw
2,759
6,194.75
past
787
infinitive
present
-ing participle
Total
Change in frequency
raw
Note: Figures exclude certain instances of have got (see discussion in text). Columns A and B represent
goodness of fit χ2 comparisons summarized in the text. Results marked “s” are significant at p < .05.
are very frequent in the present tense category, comprising around 24 percent of
examples; examination of a 10 percent random sample showed that a majority are
stative or ambiguous, so all instances were excluded from the counts.4 Occurrences
with got were far less frequent in the other inflectional categories; all examples
were examined, and stative and ambiguous ones excluded from the counts (necessary only for the past perfect).
Frequencies of the perfect (normalized per million words, “pmw”) were then
compared for LLC and ICE-GB. The results show that the perfect auxiliary falls in
frequency by 10.59 percent across the two subcorpora (Table 1, Total row).5
However, Table 1 also shows that not every category behaves in the same manner. Past and infinitival forms fall by around 34 percent and 30 percent, respectively, whereas the present (by far the most frequent of the forms) is stable and
indeed slightly increases in real terms. We report two distinct series of chi-square
tests: in Column A we compare the distribution of each term (present, past, etc.)
with the total number of words; in Column B we compare each term relative to
the trend of the overall set of perfect auxiliaries.6 Note that the slight percentage
4 This does exclude some genuine perfect examples. An alternative calculation excluded
only stative and ambiguous examples involving got for the present tense category,
based on estimations from the random samples. This produced very similar results to
those in Table 1, since few examples are clearly non-stative.
5 Note that the data reported in this chapter are based on a revised version of DCPSE
prepared at the Survey of English Usage.
6 To be more precise, in Column A we carry out a goodness-of-fit χ2 test (Sheskin 1997:
95) for the overall change (in the Total row) and for each individual subcategory
(present, past, etc.) against the number of words in the corpus. This evaluates whether
the observed percentage change is significant (i.e. significantly different from zero; see
Figure 2). Column B uses the same test against a perfect auxiliary baseline.
16_Nevalainen_Ch15.indd 203
6/14/2012 2:37:33 AM
OUP UNCORRECTED PROOF – REVISES, 06/14/12, NEWGEN
204
rethinking evidence
Present
Past
Ininitive
-ing participle
Total
–34%
–30%
–18%
–11%
20%
10%
+1%
0%
–10%
–20%
–30%
–40%
–50%
Figure 2. Changes in pmw frequencies of perfect auxiliaries from LLC to ICE-GB (Table 1
“percent” column) with error bars for p < .05. Where an error bar does not cross the zero axis,
the change is statistically significant (cf. Table 1 Column A). This type of visualization displays
the size of the result (i.e. the column height) and our confidence in it. For example, we are 95
percent certain that the “past perfect” category falls by between −25 percent and −43 percent.
increase (0.75 percent) of the present tense category is not considered significant
compared with the number of words (Column A), but it does differ from the overall
pattern (Column B). The figures for the -ing participle are small and neither result
is significant. Figure 2 displays the changes in pmw frequencies, with 95 percent
confidence intervals depicted by error bars.7
For the present perfect, our results for spoken English in DCPSE can be
compared with those of Hundt and Smith (2009) for printed written English
in the BROWN quartet of corpora. They find a slight decrease in overall pmw
frequency, which is not statistically significant, for both BrE and AmE, from
the 1960s to the 1990s (with the frequency in AmE starting lower and remaining
significantly lower than in BrE). They also consider the relative proportions of
present perfect and morphological past tense, and again find a pattern of overall stability. Further analysis of our data gives a contrasting result: the present
perfect actually shows a significant proportional increase (around 12 percent)
against the morphological past tense.8 Thus, the present perfect deserves further investigation in terms of possible changes in use and distribution.9 Genre
variation is likely to be relevant here (cf. Gries 2006); Hundt and Smith (2009)
report some differing diachronic trends in pmw frequency for particular written
7 These were computed using the Newcombe (1998) proposed interval for the difference
between two proportions, which is based on the Wilson score interval. This is a more
precise method than traditional Gaussian error bars.
8 For further details, see Bowie, Wallis, and Aarts (forthcoming), which also reports a
significant proportional decline against the morphological past tense for both the past
perfect and the category modal+perfect infinitive.
9 An interesting narrative use of the present perfect has been reported for recent BrE,
although it may represent a longer standing nonstandard use (Walker 2008).
16_Nevalainen_Ch15.indd 204
6/14/2012 2:37:33 AM
OUP UNCORRECTED PROOF – REVISES, 06/14/12, NEWGEN
change in the english infinitival perfect construction
205
genres. As the two parts of DCPSE consist of material in a range of matched
spoken genres, it is possible to explore further by comparing diachronic trends
in these genres.
Here, however, our main concern is the substantial declines observed for the
past and infinitival subtypes, in marked contrast with the findings for the present
perfect. American influence may be a contributing factor in these declines. Elsness
(1997) reports data on the past perfect and infinitival perfect in printed English,
showing that proportions of both (within the set of past-referring forms) are significantly lower in American than in British contemporary material, having fallen
in AmE since 1800.10 However, he finds the same pattern for the present perfect,
which is not declining in our data. Two factors may help to explain the contrast.
First, as noted above, the present perfect differs from the other perfect subtypes
in having a specialized pattern of use, involving an orientation toward present,
as well as past, time. Second, the other subtypes are much less frequent than the
present perfect, and less frequent items may be more likely to suffer loss than more
frequent ones (see e.g. Leech et al. 2009: 90, 269–70). The next section discusses the
infinitival perfect in more detail (on the past perfect in DCPSE, see Bowie, Wallis,
and Aarts, forthcoming).
3. The infinitival perfect in DCPSE
The infinitival perfect occurs in two main kinds of context: a bare infinitival
construction with a preceding modal auxiliary (we should have brought Dilys
along), or a to-infinitival construction (she seems to have been far less tired).
Several FTFs were constructed for these contexts. Figure 3 shows an FTF (with
left-to-right branching) to retrieve examples occurring within a VP after a
modal auxiliary. In the corpus, a VP consists of the main verb and any preceding auxiliaries, with intervening material such as adverb phrases included.
Intervening material is allowed for in the FTF by choosing the setting “next
child: after” rather than “next child: immediately after”, shown by the white
arrow (so including examples like might quite well have died in childbirth, and
even that must I fancy from the way he played it have come back onto him very
very sharply).
For the modal context, a second FTF retrieved additional examples where the
modal auxiliary preceded the subject and was therefore separated from the VP
(in interrogatives such as How old would you have been?). Several further FTFs
were used to find to-infinitival examples occurring in structures parsed as involving “semi-auxiliaries” (e.g. be supposed to, have to, seem to) and those parsed as
10 This is based on our calculations combining figures provided by Elsness for several
different constructions involving these perfect subtypes (1997: 104, 267–68).
16_Nevalainen_Ch15.indd 205
6/14/2012 2:37:33 AM
OUP UNCORRECTED PROOF – REVISES, 06/14/12, NEWGEN
206
rethinking evidence
Figure 3. FTF for a perfect infinitive auxiliary following a modal auxiliary under a VP
clauses introduced by “particle” to (e.g. What would you claim to have achieved
specifically?).
The searches showed that, across the corpus, the great majority of examples
of the perfect infinitive (88 percent) occur following a modal auxiliary. A decline
in frequency has been observed for the modals themselves in studies of recent
change (e.g. Leech et al. 2009; Aarts, Wallis, and Bowie, forthcoming). In DCPSE,
modal auxiliaries (which total 14,316) decline in frequency, as a proportion of
words, by 6.4 percent from the earlier subcorpus to the later one (the result is significant at p < .05). This raises a question concerning the decline observed in the
infinitival perfect: Is it simply due to a decline in this major context of potential
occurrence? This was tested by (i) using the same FTFs as described above, but
with the node “AUX(perf, infin)” omitted, to quantify potential contexts; and (ii)
calculating the proportions of such contexts in which a perfect infinitive occurs.
The results for modal and to-infinitive contexts are shown in Tables 2a and 2b,
respectively.11
These results show that the proportion of perfect infinitives has fallen significantly within both kinds of contexts. Therefore, the overall decline in frequency
of the infinitival perfect is not attributable solely to the decline in the frequency
of modal auxiliaries, but involves independent trends of decline within possible
contexts of occurrence. The decline within the to-contexts is particularly steep, but
involves much smaller numbers overall.
This does not in itself explain the observed decline, as the presence and absence
of a perfect infinitive in these contexts cannot in general be considered alternative choices for expressing very similar meanings. This is evident from pairs such
as he may be in London versus he may have been in London, or he is believed to
be in London versus he is believed to have been in London, where there is a clear
11 The total of “modal contexts” is slightly lower than the total of all modal auxiliaries,
because the context FTFs exclude instances of modals with no associated VP node,
as in tag questions or elliptical utterances like could you. A similar result is obtained
if the set of all modals is chosen as the basis for comparison. Note also that there are
six perfect infinitives from the total in Table 1 that are not accounted for by the FTF
contextual searches.
16_Nevalainen_Ch15.indd 206
6/14/2012 2:37:34 AM
OUP UNCORRECTED PROOF – REVISES, 06/14/12, NEWGEN
change in the english infinitival perfect construction
207
Table 2a. Changes in the proportion of perfect infinitives in modal contexts in
the LLC and ICE-GB components of DCPSE
perfect infinitive
no perfect
infinitive
Total
χ2 score
LLC
561 (7.37%)
7,050
7,611
% change =
ICE-GB
374 (5.79%)
6,081
6,455
−21.39 (c.i. ± 11.10)
Total
935 (6.65%)
13,131
14,066
14.00 s
Note: The result is significant (“s”) for p < .05.
Table 2b. Changes in the proportion of perfect infinitives in to-contexts in the
LLC and ICE-GB components of DCPSE
perfect infinitive
no perfect
infinitive
Total
χ2 score
LLC
87 (1.33%)
6,447
6,534
% change =
ICE-GB
40 (0.64%)
6,201
6,241
−51.86 (c.i. ± 25.63)
127 (0.99%)
12,648
12,775
15.47 s
Total
Note: The result is significant (“s”) for p < .05.
temporal contrast. However, there are instances in the data where a non-perfect
variant would differ in meaning only subtly, at most:
(1) you know very well that your Party would have had to have done
something uh if it had come back to power (DL-D04 #41)12
(2) there were to have been four greys in the field but the only one left is
Marche d’Or three (DL-F04 #54)
(3) well I’d like you to have found out please (DL-H01 #130)
(4) he’s forty odd I would have thought (DI-B49 #53)
In (1) the second perfect seems superfluous, as no further anteriority is involved; of
12 such “superfluous double perfect” examples in the corpus, 10 are from LLC. Such
uses have been condemned by prescriptive grammarians from the eighteenth century onward (Molencki 2003). In (2) (from a horse-racing commentary), where past
tense were expresses anteriority, the perfect seems to convey non-actualization of
the situation, but as this is clear from the context it could have been left unexpressed
(there were to be four greys, more plausible with stress on were). In (3), which conveys
a directive, the perfect expresses a “past-in-future” interpretation (the context suggests that the future reference point is the next court hearing), but again this could
have been left unexpressed. In (4) the formulaic expression I would have thought is
12 The prefixes “DL” and “DI” indicate examples from the LLC and ICE-GB subcorpora
of DCPSE.
16_Nevalainen_Ch15.indd 207
6/14/2012 2:37:34 AM
OUP UNCORRECTED PROOF – REVISES, 06/14/12, NEWGEN
208
rethinking evidence
used as a “hedge”; in most examples with such formulae, the perfect seems merely
to add to the tentativeness expressed in I would think, perhaps because locating an
opinion in past time suggests a readiness to revise it.13 Variants of the formula {I/you/
one} {would/should} have {thought/said} are quite numerous in the corpus (70, allowing for negative and reduced forms of the auxiliaries; 47 of these are in the LLC).14
In other instances there is a possible non-perfect variant with a preceding
morphological past tense expressing past time:
(5) and apart from that I mean my results are supposed to have come out
today (DI-C03 #237)
(6) well he’s lucky to have got an extra hour in (DL-A07 #41)
In (5) we might expect instead my results were supposed to come out today (cf.
Collins 2009: 81–82), while (6) seems little different in meaning from he was lucky
to get an extra hour in.
The examples discussed above suggest there is some leeway for the use of nonperfect variants instead of constructions with a perfect infinitive, so there may be an
increasing tendency to simplify verb phrases where possible. However, this leeway
seems to apply only in a restricted set of instances; beyond these, non-perfect alternatives would involve quite different structures (such as perhaps I missed something
instead of I may have missed something), making it hard to identify a determinate set
of alternatives. This may require a broader ranging investigation of the expression of
modality in combination with past time reference. This area of English involves considerable complexity in form–meaning mappings, which may lead to variation and
instability as speakers reanalyze the mappings. Depraetere and Reed (2006: 287) note
the need for more research on the temporal interpretation of modal utterances in
English; this should include research from variationist and diachronic perspectives.
4. Conclusion
Our corpus data for spoken BrE reveals contrasting trends in different subcategories of the perfect construction over recent decades: the past and infinitival
perfects decline, while the present perfect is stable in pmw frequency and indeed
13 However, in a minority of instances the perfect is not omissible. For example, one
speaker, having expressed surprise to hear that it is raining, says I would have thought
it was too cold to rain. Here I would think it is/was too cold to rain would not be
appropriate, as the evidence forces revision of the opinion.
14 These perfect and non-perfect formulae can also be considered as a separate set (n =
222). Variants of the non-perfect formulae I would think/say are also more frequent in
LLC than in ICE-GB; the proportion of formulae which are perfect is slightly higher in
LLC (32 percent vs. 30 percent), but the difference is not significant. Removing the set
of formulae from Table 2a makes little difference to the results.
16_Nevalainen_Ch15.indd 208
6/14/2012 2:37:34 AM
OUP UNCORRECTED PROOF – REVISES, 06/14/12, NEWGEN
change in the english infinitival perfect construction
209
increases proportionally against the past non-perfect. Factors behind the declines
may include American influence and a tendency to simplify the VP. The different
pattern for the present perfect may be due to its much greater frequency and to its
semantic specialization (involving reference to present, as well as past, time).
Closer examination of the infinitival perfect shows that it most frequently
occurs following a modal auxiliary. Modals have themselves fallen in frequency;
however, by taking into account the frequencies of possible contexts of occurrence
(i.e. modal and to-infinitival structural contexts), independent trends of decline of
the perfect infinitive within these contexts can be shown.
This study has shown the importance of considering changes in a linguistic
category like the perfect in relation to its interaction with other categories like
morphological tense and modality. The interaction of categories is likely to be
important in change processes, especially in areas where form–meaning mappings
are complex. Our study has also highlighted the need to consider the relative frequencies of these interacting categories. The investigation of such complexities is
facilitated by use of a parsed corpus like DCPSE with a flexible means of searching
for structural patterns.
References
Aarts, Bas, Gerald Nelson, and Sean Wallis. 1998. ‘Using Fuzzy Tree Fragments to
Explore English Grammar’. English Today 14: 52–56.
Aarts, Bas, Sean Wallis, and Jill Bowie. Forthcoming. ‘Proiling the English Verb
Phrase over Time: Modal Patterns’. In Corpus Linguistics and the Development of
English, ed. Merja Kytö, Irma Taavitsainen, Claudia Claridge, and Jeremy Smith.
Cambridge: Cambridge University Press.
Biber, Douglas, Stig Johansson, Geofrey Leech, Susan Conrad, and Edward Finegan.
1999. he Longman Grammar of Spoken and Written English. London: Longman.
Bowie, Jill, Sean Wallis, and Bas Aarts. Forthcoming. ‘he Perfect in Spoken British
English’. In he Verb Phrase in English: Investigating Recent Language Change with
Corpora, ed. Bas Aarts, Joanne Close, Geofrey Leech, and Sean Wallis. Cambridge:
Cambridge University Press.
Collins, Peter. 2009. Modals and Quasi-Modals in English. Amsterdam: Rodopi.
Depraetere, Ilse, and Susan Reed. 2006. ‘Mood and Modality in English’. In he
Handbook of English Linguistics, ed. Bas Aarts and April McMahon, 269–90.
Malden, MA: Blackwell.
Elsness, Johan. 1997. he Perfect and the Preterite in Contemporary and Earlier English.
Topics in English Linguistics, 21. Berlin: Mouton de Gruyter.
Fischer, Olga, and Wim van der Wurf. 2006. ‘Syntax’. In A History of the English
Language, ed. Richard Hogg and David Denison, 109–98. Cambridge: Cambridge
University Press.
Gries, Stefan h. 2006. ‘Exploring Variability within and between Corpora: Some
Methodological Considerations’. Corpora 1: 109–51.
Huddleston, Rodney, and Geofrey Pullum. 2002. he Cambridge Grammar of the
English Language. Cambridge: Cambridge University Press.
16_Nevalainen_Ch15.indd 209
6/14/2012 2:37:34 AM
OUP UNCORRECTED PROOF – REVISES, 06/14/12, NEWGEN
210
rethinking evidence
Hundt, Marianne, and Nicholas Smith. 2009. ‘he Present Perfect in British and
American English: Has here Been Any Change, Recently?’ ICAME Journal 33:
45–63.
Leech, Geofrey, Marianne Hundt, Christian Mair, and Nicholas Smith, eds. 2009.
Change in Contemporary English: A Grammatical Study. Cambridge: Cambridge
University Press.
Molencki, Rafał. 2003. ‘Proscriptive Prescriptivists: On the Loss of the “Pleonastic”
Perfect Ininitive in Counterfactual Constructions in Late Modern English’.
Linguistic Insights: Studies in Language and Communication 7: 175–96.
Nelson, Gerald, Sean Wallis, and Bas Aarts. 2002. Exploring Natural Language:
Working with the British Component of the International Corpus of English.
Amsterdam: Benjamins.
Newcombe, Robert G. 1998. ‘Interval Estimation for the Diference between
Independent Proportions: Comparison of Eleven Methods’. Statistics in Medicine 17:
873–90.
Sheskin, David J. 1997. Handbook of Parametric and Nonparametric Statistical
Procedures. Boca Raton, FL: CRC Press.
Smith, Nicholas, and Geofrey Leech. Forthcoming. ‘Verb Structures in TwentiethCentury British English’. In he Verb Phrase in English: Investigating Recent
Language Change with Corpora, ed. Bas Aarts, Joanne Close, Geofrey Leech, and
Sean Wallis. Cambridge: Cambridge University Press.
Svartvik, Jan, ed. 1990. he London-Lund Corpus of Spoken English: Description and
Research. Lund: Lund University Press.
Walker, Jim. 2008. ‘he Footballer’s Perfect: Are Footballers Leading the Way?’ In
he Linguistics of Football, ed. Eva Lavric, Gerhard Pisek, Andrew Skinner, and
Wolfgang Stadler, 295–303. Tübingen: Narr.
16_Nevalainen_Ch15.indd 210
6/14/2012 2:37:34 AM