Bibliotheca Arnamagnæana - 01.06.1999, Page 300
280
Part Two
times, and we have already seen de Vries and Kuhn offering different
explanations for the subsequent increase.18 This part of the graph is of no
relevance to de Vries’s view on the consequences his findings have for
the dating of Eddie poetry, however, as this hypothesis is based on the
latter half of the graph. Starting from a minimum in the period
1050-1100, the graph shows an increase in the following two periods,
after which it once again decreases. This peak in the curve of the graph
suggests to de Vries that the figures reveal a literary heathen revival
from 1150 onwards, which once again made it possible to compose
Eddie mythological poems after a gap of about 150 years. There is no-
thing startling about this conclusion, but the point is that de Vries’s
figures make it possible to localize a definite period in the history of Old
Norse poetry in which mythological kennings were so sparsely used that
it can be surmised that mythology must have been banned also from
Eddie poetry.
One of de Vries’s reviewers praised him for basing his conclusions on
percentages instead of raw numbers (Hollander 1935: 90). Calculation
of percentages has a serious drawback, however, in that it gives no less
importance to figures based on small numbers than to figures based on
large numbers, and therefore it should be avoided wherever possible, not
least in this case where the basic figures (number of stanzas in each
period) vary from 70 to 884. Another question not raised by de Vries is
whether the fluctuations in the occurrence of OSinn-kennings may
simply be due to random variations in his material. We can never be sure
that the sample constituted by the material handed down to us gives a fair
representation of historical reality. But there are methods for testing
the probability that this is the case. A simple way of testing the nuil hy-
pothesis, i.e. the possibility that the observed fluctuations are due to ran-
dom variation, is to apply to the material a chi-square test. In this case it is
immediately evident that the very marked rise in the period around the
conversion is significant, but less so for the latter part of the graph,
which we are concemed with. To test the significance of the peak, we
18 The two explanations are not incompatible, however, as it is theoretically conceivable
that a more fervent religious feeling may have raised the number of mentions of OSinn in
some instances, while a growing indifference at the same time may have lifted the taboo
on certain religious kennings on the other hånd, also leading to an increase. My intuitive
guess is that a composite explanation of this kind is no less realistic than the more straight-
forward alternatives.