[Corpora-List] Re: Phrasal verbs

[Sampo Nevalainen]

At 18:16 13.5.2003 -0500, FIDELHOLTZ DOOCHIN JAMES LAWRENCE wrote:
>Brett,
>Now, 2000 per M boils down to 1 per 500 words of text, which, since we are
>talking about 2-word units is 2 per 500, or 1 per 250 words, that is, on
>average one per page (OK, let's not exaggerate: one every other page).  I
>don't know how much work you have done with frequency, but this seems to
>me to be a very high frequency of occurrence, well worth stressing in an
>ESL class.  By contrast, check how often in, say, a 10-page article you
>encounter the passive.  Probably not more than 5 times, I would guess,
>that is, more or less the same frequency, and I doubt you would question
>teaching the passive.
>Jim

Sorry for messing into this, but I just got interested in the problem of
counting collocations. I don't know how the proportion of 2000 per million
( or 1 per  500) was calculated in the Longman Grammar of Spoken and
Written English, but I think one cannot derive 1 per 250 from these
figures, since phrasal verbs are treated and should be treated as "one
word" -- a proportion of 1 per 250 would count in one phrasal verb two
times. That is, the frequency of a phrasal verb is obviously not the
cumulative frequency of it's elements. However, if the elements of phrasal
verbs are regarded as separate words in counting the total number of words,
but as one word in counting the relative frequencies of phrasal verbs, then
I don't really know what we get as a result. To make this more complicated,
it might be noted that there may be more than two elements in a phrasal
verb (for example 'add up to' is inseparable in 'The bills add up to $50' ;
cf. 'add up', which is separable - 'add them up') Now, this thing gets
really messy... even without my crooked English :-) Could somebody tell me
how the proportions of phrasal verbs and other collocations are counted, or
am I totally lost? (As a concrete example, should we, knowing that there
are 500 occurrences of 'come down with', imagine it as one word by
subtracting the proportion of the "tied" element(s),  in this case, 2 * 500
= 1000, from the total number of words -- and do this procedure for every
single phrasal verb type in the corpus before we can count their relative
frequency..?)

sincerely,
sampo