Phraseology and Culture in English

Appendix I: 
1. Frequencies for ten neutral collocations in the British National 
Corpus, the COBUILD Collocations Database and selected top-
level domains on the World Wide Web 
The BNC contains c. 100,000,000 words, the COBUILD database slightly 
over 200,000,000. Comparing the frequencies for these two corpora shows 
that, as expected, the figures for COBUILD are consistently higher, but 
also that there still is considerable fluctuation in the frequency of occur-
rence of individual collocations. The figures for the individual top-level 
web domains need to be interpreted in this light. They give a rough idea of 
the proportion of the various domains relative to each other, and they allow 
a rough estimate of the amount of material looked at in comparison to the 
100,000,000 words of the BNC. Thus, the .edu and .uk domains appear 
broadly comparable in size, and so do the .au and .ca ones, which seem to 
contain roughly a fourth of the material found in the two bigger ones. 
While it is fairly safe to determine rough proportions, it is much more risky 
to calculate approximate numbers of words. For example, by extrapolation 
from the BNC, the 7,650 instances of deep breath in the Australian web 
material would indicate a size of c. 1.34 billion words. Performing the same 
calculation on early age, on the other hand, we would arrive at the rather 
different estimate of 5.5 billion words. In view of such fluctuation, esti-
mates should not be based on individual collocations but on aggregate fre-
quencies.

---

460
Christian Mair 
Table 2. 
Distribution of established collocati
onal m
arkers of Britishness on the 
W
eb (figures = rounded percentages) 
sum of all