Íslenskt mál og almenn málfræði - 01.01.1985, Page 132

130
Christer Platzack
1. Frequency studies of word order pattern — some methodological
considerations
There are many things to consider before starting to count the fre-
quency of some grammatical category in a given text. As in all statisti-
cal investigations, there is e.g. the question of the representability of
the sample. I will not consider such general questions here, only refer
to Kohonen (1974), where several problems in connection with the
statistical investigation of word order pattems in texts are discussed.
A general problem for all kinds of historical linguistics is to find
reliable editions of the texts investigated. Sigurðsson bothers a lot
about this problem — it is obviously an important piece of informa-
tion he provides us with when he demonstrates that the íslenzk fornrit
edition of the Sagas does not always give us the correct word order
found in the manuscript underlying the edition. Occasionally, words
are inserted or deleted, and sometimes the word order is changed. The
existence of such shortcomings obviously forces us to be extra careful
when we consider results based solely on the íslenzk fornrit edition.
To be able to perform a quantitative study, we have to have a
grammatical theory according to which the text can be analyzed. This
theory must define the categories of interest as unambiguously as
possible — naturally, we do not have much use for frequency figures if
we do not know what they represent. Sigurðsson and Kossuth, as well
as most scholars doing quantitative linguistics, use some kind of tradi-
tional school-grammar with concepts like subject, finite verb, object,
clause, etc. Concepts like these are familiar enough, but since they
have been defined in somewhat different ways by different scholars, we
cannot use them uncritically. Rather, we have to state (as Sigurðsson
does) as exactly as possible how we are going to use them in our in-
vestigation.
Naturally, when every scholar has to make his own decision how to
define the basic categories, there will usually be some differences
between different investigations, which sometimes may make the
results hard to compare. The awareness of this problem has led a
group of Swedish scholars interested in doing quantitative linguistics to
publish two manuals for grammatical analysis of spoken and written
Swedish, Loman & Jörgensen (1971), and Teleman (1974). It has been
shown in several studies that these manuals can be adapted also to