The changing verb phrase in present-day British English
The core modals
The items included here as core modals are can/could, will/would, shall/should, may/might, and must (with the paired items indicating present and past tense counterparts).
- Have the core modals, as a group, declined in frequency in spoken British English from the 1960s to the 1990s, as has been found for the written language (e.g. Leech et al. 2009)?
- What are the patterns of change for individual modals over this period?
- What are the patterns of change across different spoken text categories (e.g. informal conversations, broadcast interviews, prepared speech)?
Methods for measuring frequency
- Modal frequency is often measured per million words. Another method we have applied is to measure frequency per tensed VP. A modal auxiliary can only occur within a tensed VP. Speakers can only exercise the choice to use a modal within this grammatical context. Therefore tensed VPs provide a useful baseline for measuring modal frequency. This method is more precise as it controls for the variation in 'tensed VP density', i.e. the variation in frequency of tensed VPs per million words. This is important, as we have found considerable variation in tensed VP density across different spoken text categories (e.g. formal conversations, broadcast interviews, prepared speech). (See findings on Verb phrase density.)
- It is also informative to consider the frequency of each modal in relative terms, as a member of the set of core modals (using the set of all core modals as the baseline).
Summary of main findings
- As a group, the core modals have declined in frequency per tensed VP from 14.20% to 13.32%, from the earlier to the later subcorpus of DCPSE (i.e. from the 1960s-70s data of LLC to the 1990s data of ICE-GB). This represents a significant decline of around 6%. (A similar result is obtained for the frequency change when frequency is measured per million words.)
- Three modals show large and significant declines: may (about 40%), must (54%), and shall (48%). Three other modals show smaller but significant declines, ranging from 7% to 14%: would, could, and should. One modal, will, shows a significant increase of 13%. No significant change is found for the remaining two items, can and might.
- Different spoken text categories show considerable variation in modal frequencies, both synchronically and diachronically. This applies both to the modals as a group and to individual modals. For overall modal frequency, four text categories show significant decreases, while only one (the small and specialised category of legal cross-examination) shows a significant increase.
- The modals may, must, and shall show the most consistent patterns of change across the text categories: each declines significantly in four or more categories (including, in each case, the largest category of informal conversations), and none increase significantly in any category.
- The two different methods for measuring modal frequency, per million words and per tensed VP, produce notably different results for change over time when the modal data is broken down by text category. This shows the importance of considering the baseline in measuring the frequencies of grammatical items such as the modals, especially when different kinds of text category are compared.
Tables and graphs: individual modals, all text categories combined
Table 1. Modals per tVP, across the two subcorpora of DCPSE, by individual modal. Truncated and negative forms are included (e.g. 'll, won't). Results in bold are significant at p<0.05.
Table 1 shows the frequencies of the modals per tensed VP in the two subcorpora of DCPSE, and the percentage changes in frequency per tensed VP from the earlier subcorpus (LLC) to the later one (ICE-GB). Column A reports the results of chi-square tests for the frequency changes per tensed VP, with figures in bold indicating significant results. For example, could declines in frequency per tensed VP from 1.57% in LLC to 1.39% in ICE-GB, and this represents a significant decline per tensed VP of around 11%.
Column B of Table 1 reports the results of a distinct series of chi-square tests which compare the change pattern for each individual modal with the pattern for the overall set of modals. This allows us to consider the behaviour of each modal in relative terms, against a baseline of all modals. Note, for example, that the increase in can is not considered significant compared with the number of tensed VPs (Column A), but this modal does significantly increase its share of the modal set (Column B).
The results in Table 1 are visually displayed in the two figures below.
Figure 1. Modals per tVP in the two subcorpora of DCPSE, by individual modal.
Figure 1 shows the frequency per tVP of individual modals in the two subcorpora of DCPSE. It can easily be seen that individual modals vary greatly in frequency. For example, can, will, and would are much more frequent than the other modals in both the earlier subcorpus (LLC) and the later one (ICE-GB), while shall is the least frequent in both subcorpora. Change over time is more clearly shown in Figure 2 below.
Figure 2. Percentage changes in frequency of the modals across the two subcorpora of DCPSE, per tensed VP and per modal. The error bars represent 95% confidence intervals. Where an error bar does not cross the zero axis, the result is significant.
Figure 2 shows the percentage changes in frequency of the modals from the earlier to the later subcorpus, both per tensed VP and per modal. May, must, and shall show large and significant declines both per tensed VP and per modal. Could, should, and would show smaller significant declines per tensed VP, but the changes relative to the modal set are not significant. (They can be described as 'typical modals' in that their behaviour over time cannot be distinguished from that of the overall modal set.) Will increases significantly both per tensed VP and per modal, while the increase in can is significant only relative to the set of all modals.
Tables and graphs: all modals together, by text category
Table 2. Modals per tVP, across the two subcorpora of DCPSE, by text category. Results in bold are significant at p<0.05.
Table 2 shows modal frequency per tVP in the two subcorpora, broken down by text category. It also shows the percentage frequency changes from LLC to ICE-GB for each text category, with the results of chi-square tests for significance. The results are visually displayed in the two figures below.
Figure 3. Modals per tVP, across the two subcorpora of DCPSE, by text category.
Figure 3 shows the frequencies in LLC and ICE-GB. It can easily be seen that there is considerable variation in modal frequency among the text categories. For example, parliamentary language shows a comparatively high modal frequency in both subcorpora, while the frequency for spontaneous commentary is comparatively low in both subcorpora. Change over time is shown more clearly in Figure 4 below.
Figure 4. Percentage changes in frequency of the modals per tensed VP, across the two subcorpora of DCPSE. The error bars represent 95% confidence intervals. Where an error bar does not cross the zero axis, the result is significant. The large increase in the legal cross-examination category is not fully displayed.
Figure 4 shows percentage changes in frequency per tensed VP from LLC to ICE-GB. Four text categories show a significant decrease in overall modal frequency: telephone conversations, broadcast discussions, broadcast interviews, and parliamentary language. Only one category, that of legal cross-examination, shows a significant increase in overall modal frequency.
A paper in preparation will provide details of change in individual modals across the different spoken text categories.
Further findings on recent change in the modals are presented in papers by Aarts, Wallis and Bowie (forthcoming 2012); Aarts, Close and Wallis (forthcoming 2012); and Close and Aarts (2010). The first of these looks at modal change in a wide range of structural patterns (e.g. declarative/interrogative, multiple-auxiliary structures); the second includes a case study of will/shall; and the third presents a case study of must and have (got) to. Versions of these papers can be downloaded from the main page for the project, The changing verb phrase in present-day British English.
This page last modified 12 June, 2013 by Jill Bowie.