Tertia comparationis in multilingual corpora 9
comparing expressions coding ‘betweenness’ in two languages. The tertium com-
parationis in this case is therefore semantic/pragmatic, rather than syntactic equi-
valence. This does not mean that semantically equivalent expressions may not also
be syntactically equivalent. Consider in this respect the pairs of sentences in (3)
and (4), taken from the Oslo Multilingual Corpus (OMC).
(3) a. Markus Torjussen takes his hand between both of his and shakes it
emphatically. (BHH1TE)
b. Markus Torjussen prend sa main entre les deux siennes et là secoue vigou-
reusement. (BHH1TF)
(4) a. And the lovely paths among the flower beds, strewn with crushed shells!
(HW2TE)
b. Et les jolis sentiers entre les plates-bandes, recouverts de sable! (HW2TF)
The pairs of sentences in (3) and (4) are both syntactically congruent, as well
as semantically equivalent. They thus conform to the requirements proposed by
Krzeszowski (1990) for primary data for contrastive analysis.
In an earlier work (Krzeszowski 1981: 123), I suggested that syntactic contrastive
studies should be performed on data restricted in the following way: a contras-
tive grammar will take as its primary data (to be assigned the status of semantic-
syntactic equivalence) the closest approximations to grammatical word-for-word
translations and their synonymous paraphrases, if such forms exist. Such a con-
straining of primary data as the basis for syntactic contrastive studies bypasses the
inherent difficulties of the proposals suggesting the use of unrestricted semantic
equivalence as the basis for comparison. (Krzeszowski 1990: 19)There is no doubt that the pairs of sentences in (3) and (4) comprise close “approx-
imations to grammatical word-for-word translation”. Indeed they are grammati-
cal word-for-word translations, though not of each other. Krzeszowski (1990: 25)
employs the term 2-text to refer to texts in either parallel or translated corpora.
The availability of multilingual corpora, such as the OMC, allows us to operate
with the concept of the 3-text, with expressions in a source language serving as
potential tertia comparationis for their translations into two other languages. This
means that we have another means of solving the problem of unrestricted seman-
tic equivalence raised by Krzeszowski. The essence of this problem is the difficulty
of identifying the exact semantic content common to a source item and its transla-
tion. What the two sets of translated items in tokens such as (3) and (4) have in
common is simply the fact that they are both translations of the same source items.
The fact that we have a tertium comparationis in the form of the Norwegian
original also means that word-for-word translations are not privileged in any way,
as they are in Krzeszowski’s approach. What is of interest is how the other two