Bibliotheca Arnamagnæana - 01.06.1999, Page 229
VII The expletive particle
209
Statistical evaluation
My point of departure is the observation that although not only Kuhn,
but philologists in general often use statistics in their work, and even
more often follow a trend of thought which is statistical in nature even if
no explicit statistics are presented, they seldom feel any obligation to re-
flect on the scope or importance of their method. The present essay is
not to be understood as a statistical work in its own right. Having no sta-
tistical training, except what is given in introductory works for linguists
(e.g. Muller 1968, cf. 1981, or Butler 1985), I am well aware of the lack
of expertise underlying these reflections. But since statistical reasoning
will be seen to be relevant in connection with a number of the methods
of dating used in discussion of the Edda, I think it is appropriate to go
somewhat further into the matter than is strictly necessary for the imme-
diate purpose. The statistical manipulation of my tables may be more
complicated than is needed for the rather common-sense conclusion
finally reached - for my purposes it does not really matter much whether
p is 0.7 or 0.8 (cf. p. 216 n. 10 below). In general I think it is important
to choose methods that are well adapted to the purpose of the work, and
too high a level of sophistication may be out of place if it has no bearing
on the results obtained. In this case, however, I have decided to retain
my tables for the sake of exactitude and clarification - and by reference
to the exploratory nature of my essay. At all events, the discipline to-
wards greater explicitness and exactitude exerted in the transformation
of data from qualitative to quantitative units remains an important gen-
eral benefit of a statistical method. In order to quantify the texts, they
will have to be treated in a well-defined and uniform manner, according
to what Charles Muller calls “norme de dépouillement” (the norm for
extraction of data).3 This is not always the case in non-statistical philo-
logical analysis. Further, it places the reasoning on a level of abstraction
where the methodical principles involved become more transparent.
Statistical method is the handmaid of philology, not a mistress exacting
supplies from philological data for her own sake; but she is a very help-
ful and intelligent handmaid.
In order to show the decrease in the use of the particle in skaldic
3 “[...] l’ensemble des régles ou des conventions qui, dans le dépouillement quantitatif
d’un texte, garantissent la Constance du traitement et de ses resultats” (Muller 1968: 142).