Notes on the Voynich Manuscript - Part 17	[1992 February 5]
-----------------------------------------

Where have all the Consonants gone?

On the basis of our current hypotheses, the Voynich text has
too few consonant occurrences and too few consonantal letters.
Here's the statistics, for the entire Currier text:

    count    permill  C/V ??

O:  10435    156	V
9:   8186    122	V
C:   7188    107	V
8:   6001     90	C
A:   5115     76	V
S:   5058     75	C
E:   4362     65	V
F:   4090     61
R:   2686     40
4:   2564     38
P:   2394     35
Z:   2270     34
M:   1717     25
2:   1005     15
N:    749     11
Q:    603      9
B:    503      7
X:    463      6
J:    350      5
V:    156      2
T:    134      2
W:    122      1
D:     98      1
3:     82      1
----------------
U:     56      0
I:     55      0
6:     53      0
Y:     33      0
K:     16      0
7:     13      0
G:     13      0
0:      7      0
H:      7      0
L:      7      0
5:      2      0

I've inserted a cutoff line at 0.1% - everything below occurs less than
one time in 1000 characters.  Now, based on our current guess, that gives
as % of text:

	52.6% vowels
	47.4% consonants

And that's assuming everything not identified is a consonant.

For a sample of English text, I found

e:   2730    127	V
o:   1717     80	V
a:   1683     78	V
t:   1681     78
n:   1504     70
r:   1469     68
s:   1409     65
h:   1401     65
i:   1315     61	V
l:   1010     47
d:    901     42
f:    743     34
u:    558     26	V
c:    519     24
w:    469     21
m:    449     20
b:    413     19
y:    344     16	V/2
g:    333     15
p:    260     12
v:    221     10
k:     84      3
j:     45      2
x:     36      1
----------------
q:     17      0
z:      8      0

And that gives

	38.0% vowels
	62.0% consonants

Yes, but English has all those consonant clusters.  True, but an analysis
of some mediaeval Latin gave:

	43.6% vowels
	56.4% consonants

[Note: the Latin in question was most of the text of the Mass, simply
concatenated.  It was all I could find at the time.]

Now, when you look at supposed romanised Voynich, the cause is clear:
an awful lot - an humungous lot - of vowel clusters, whether long vowels,
diphthongs, or just EIEIO chanting is unclear.  Why should that be?  I
have a speculation, maybe for another note.

[Note: the vowel/consonant ratio is a property of three things: the
language, the script, and the rules of orthography.  For example,
if the language is Russian, you get a much higher vowel ratio with
the Cyrillic script than with romanised text, because the former
uses single letters where the latter must use clusters of 2, 3 or
even 4 consonants.  As another example, Sanskrit in conventional
Devanagara script gives a low vowel ratio, because the most common
vowel ('a') is usually elided in traditional orthography.  Really
weird ratios can occur when a script devised for one language is
used to transcribe a very different language - a point discussed
in a later note.]

But a bigger trouble, to my mind, is that there aren't enough symbols
to make up a decent set of consonants.  If we set the cutoff at the
1 per mill, there are just 24 Voynich letters worth counting.  We
suspect at least 5 to be vowels, and 4 more (the cXXt) to be some
form of fusion, and that leaves only 15 genuine consonants.  At the
same cutoff, English has 19 (counting Y), and Latin has 17 (all its
consonants occur at least at the 0.2% level), or 18 if you recall
that C does double duty as kappa and chi.

Again, one explanation is that many consonantal sounds are written
with two letters.  In English, for example, we could use ph, zh, bh,
ks, kw for f,j,v,x,qu so reducing the alphabet to 14 consonantal letters.
But I don't see much sign of that in Voynich; apart from 8, S(ct) and
Z(c't), few consonants seem to like to cluster.  And the sheer frequency
counts suggest that letter pairs do not form single consonants - the
reason English scores so low on vowels is precisely that a lot of our
consonantal sounds take several letters; with Voynich the inference
is the opposite, that their consonants are single letters but their
vowels are multiple letters.

So, if this is a European language, where have those consonants gone?

(Plump nymphs picked them every one?)

Robert

[Note: and, again - if this is a European language, which language
might the phenomena imply?  Yet again, the lack of a large sample
of machine-readable texts in diverse languages brought the work to
a halt.]