Studia Islandica - 01.07.1963, Blaðsíða 97

95
veal definitive individual marks af the author. But in spite of these
rather unfavourable auspices the experiment was made. Though the
result might be negative, it would be a result.
The inquiry makes use of two separate methods, independent of
each other. First, in a large number of texts the frequency of certain
words, characteristic of Laxdœla, has been registered. This section of
the work has profited much from the discussion in Alvar Ellegárd’s
treatise A statistical method for determining authorship, The Junius
letters 1769—1772, Göteborg 1962 (= Gothenburg Studies in Eng-
lish 13). Secondly, certain more infrequent words, named here pair
words, have been arranged and studied in accordance with the me-
thod developed in the author’s previous paper on Snorri Sturluson
and Egla (Studia Islandica 20).
3. The frequency of a number of adjectives (adverbs) and ab-
stract nouns. (Pp. 26—45). The frequency of certain words — both
the very common ones and the comparatively rare ones — can fluc-
tuate considerably from one saga-text to another. One author has a
preference for certain words, another uses them more seldom than a
colleague of his or the majority of his colleagues. But in order to have
a real measure of such divergences one has to establish average fre-
quencies on a broad textual basis.
For the present purpose six texts were chosen as control material,
all of which had been used in the Snorri-£g/a investigation: Snorri A
and B (= Heimskringla), Egla, Eyrbyggja, Njála and Grettla. To get
a better balance between family sagas and Kings’ sagas this material
was completed with another King’s saga, the monk Oddr Snorrason’s
saga of Ölafr Tryggvason, the A-version in Finnur Jónsson’s edition
of 1932. This makes in all a text material of ca. 529000 words, taken
to represent an average of saga-language. It does not include either
of the two sagas to be particularly tested: Laxdœla and Knýtlinga.
Then Laxdœla was searched for words more frequent in this saga
than in the control material — a very time-consuming procedure,
though considerable restrictions were practised in the choice of words.
Only adjectives (adverbs) and abstract nouns were taken into account,
as those word categories seem to reveal the most characteristic traits
of the individual sagas. Furthermore, the condition was made that each
of the words included should be at least twice as frequent (in relation
to the text volume, of course) in Laxdœla as in the control material
— in order to have the word series as representative as possible.
Naturally Knýtlinga could have been chosen as starting-point and
the word-list composed on the basis of that saga, since Laxdœla and
Knýtlinga participate on quite equal terms in the comparison. But every-