-3-
4xii50
Estimate of sign-groups available for comparison.
Length A B
2-letter 165 x .85 : 140 400 x .85 : 340
3 245 x .72 : 176 1040 x .72 : 750
4 85 x .61 : 52 1060 x .61 : 660
5 5 x .52 : 2½ 150 x .52 : 78
6 11 x .44 : 5
7 11 x .37 : 4
TOTAL: (500) 380½ (2700) 1827
First let's work out the probable number of correspondences between
these two samples of the vocabulary on a completely random basis. I cal-
culate this to be as follows:
2-letter groups 18.5 probable correspondences
3 ············· 11.0
4 ············· .6
5 ············· negligible
TOTAL: 30.1
The surprising fact is that, if the figures are roughly correct
and the A and B samples are regarded as selections out of a single
homogeneous 'vocabulary' of finite size (whether nouns or proper names
or a stable mixture of the two), the actual correspondences found are
less than half the number which one would expect even if the values of
the signs had been completely jumbled in the change-over from one syl-
labary to the other. This result is so odd as to be almost contrary to
nature !
(vii) If one assumes a finite vocabulary, the number of correspondences
to be expected is quite large:-
Size of "common vocabulary" Probable number of correspondences
2,500 ·························· 297
5,000 ·························· 148
7,500 ·························· 99
10,000 ·························· 74
What the size of this statistical "common vocabulary" is as between
Knossos and Pylos I don't know, but it's a sum one could quickly do
from data which you probably already possess, by the formula:-
Total vocabulary = Pylos range x Knossos range,
Number common to both
for example: (P) 750 x (K) 1200 =2000
------------------
450
Now I still resist the solutions of assuming a radical change in
phonetic values between even the A and B signs outwardly identical,
or a radical change in language (the effects of which might in any case
be less formidable than expected if we assumed that a proportion of
the sign-groups are place-names, or refer to minor officials and arti-
sans of a menial and presumably fairly constant population). So we've
got to find ways of mitigating the conclusions from your figures (you
might say, of cooking them!) :-
(viii) The number of HT sign-groups which are distinct, complete and free
of A-peculiar signs doesn't seem to me, on going through my lists, to
exceed about 205; and I suggest that your B figures also ought perhaps
to be reduced in the same proportion to yield a fair comparison.
This would bring the total correspondences likely on a random chance
down to 10.6 sign-groups. So we're slightly better than 'random' now,
and no longer contrary to nature.
The correspondences for finite common vocabularies would come down
to:-
Size of "common vocabulary" Probable number of correspondences
2,500 ······················· 110 60
5,000 55 32
7,500 37 23
10,000 28 19