Bibliotheca Arnamagnæana - 01.06.1999, Blaðsíða 233

VII The expletive particle
213
Having ranked the skalds according to how apt they were to use the
particle on the one hånd, and their age on the other, the two lists may be
compared by help of Spearman’s coefficient of correlation, which pro-
vides a measure of how closely related the two lists are.* * * * 5
Some problems are involved, however, in establishing the real rank of
each skald as regards his use of the particle. In particular, it is obvious
that if skalds who left only small fragments were included, sheer coinci-
dences might influence the result. If in a fragment of one stanza there
happened to be one or two occurrences of the particle, this fragment
would top the list with 10 or 20 particles per 10 stanzas, even if it could
be shown that no existing poem has anything like this frequency. To this
problem I do not think there exists any obvious solution. One remedy to
this difficulty would be simply to exclude smaller fragments, and to a
certain extent I have done so by excluding “minor” skalds, defined as
skalds by whom fewer than 80 verses have survived.6 A more sophisti-
cated method has been proposed by Charles Muller, however. Instead of
simply calculating the bare frequencies of a given linguistic phenome-
Bår5r 1.4, Hfr II 2.3, III 10.3, 19.1, 21.2, 24.4, V 2.5, 5.7, 13.7, 25.4, 28.5, BårSardr 1.1,
Anon X I B 7.7, Bjark 1.8, Anon X II B 4.2, 9.3, GDropl 1.4, 5.4, Gunnl II 1.1, Edå5 4.7,
Hokr 2.6, PKolb I 3.3, III 1.7, IV 5.3, Sigv II 4.4, 9.7, III 3.3, 19.3, VII 5.8, 8.5, XII 17.6,
26.2, XIII 6.7, 8.3, 17.2, 27.2, Porm II 1.3, 1.5, 5.3, 11.1, 19.6, Ott II 7.5, 10.8, 14.3, 16.4,
17.5, Skuli 15.1, EPver 1.8, Grett 11.8, Giz 11.3, Ploft III 9.5, PSær IV 3.3, Am 12.3, II 9.3,
11.7,13.8,14.8, III 8.1, HharS 3.3, PjoOA 114.4,16.3, II 3.1, III 28.3, IV 1.6,2.3,Bq1v 1.3,
SnH II 6.2, Valg 6.3, 9.1, BjH 3.8, 5.7, 6.8, 7.3, Pfagr 6.7, 8.1, Pfisk 3.2, Stufr 8.6, Ofeigr
2.6, PSkall 2.4, Sveinn 1.3, Darr 6.8, 7.4, 8.2, Anon XI lv 6.3, Drv 9.8, 11.6, Mark III 1.4,
GullåsP II 1.3, Sjors 3.5, ESk VI 62.4, 64.2, VIII 2.4, XI 4.3, 10.3, XIII 5.2, Kolli 5.3, Hl
la.5,22b.4, HSt 16.2,8.3,8.6, fsldr 23.5, Gmlkan 13.6,4.3, II 2.8,3.3,16.1,57.7,58.4, Od
6.8.7.7.13.6, 14.8, Ekul 13.1, Anon XIIC 36.7, Sol 8.5,53.2, Pul 12.5,3.6, IV ff 1.8.
5 Spearman’s coefficient of correlation is appropriate when the variables compared are
ordinal, based on ranking. It is denoted by p, lying between + 1 for perfect positive cor-
relation and - 1 for perfect negative correlation, and it is calculated according to the for-
mula below, in which d is the difference between the ranks and N is the total number of
observations:
P= 1 -
61cf
N (N2- 1)
(cf. Muller 1968: 119-20; 1981: 179-80; Butler 1985: 145-48).
6 For this choice I have no theoretical foundation, but by way of experiment I have found
that this is somehow a critical number. As soon as considerably smaller fragments are in-
cluded, the results become unstable. If “very minor” skalds, who left as few as 40 verses
are included in the material, the positive correlation between age and particle frequency is
greatly reduced (p = 0.55, against p = 0.79 for the actual corpus). If it is expanded to “very
major” skalds by whom at least 100 or 200 verses have survived, on the other hånd, the
correlation is basically unchanged.