On 4/04/2010 1:39 AM, Dennis King wrote:
> A mathematical analysis from the University of Exeter:
The first stage of the analysis seems to be depend on measuring the degree
to which certain symbols occur more often than others (their relative
frequency: p 4).
So, if the symbols were chosen more or less randomly, they would tend to
occur more or less at the same rate. However, as the authors say at p 2, it
is "admittedly unlikely" that the symbol stones are random in nature "since
they appear to have been created for a pourpose." So the resuklts of the
first stage of the analysis, that they are indeed not, come as no surprise.
The next stage seems to involve looking at the extent to which repetition
occurs within a corpus.
So, with morse code, their is a high degree of repetition: you can guess
with some confidence that the next character in a text will be either a dot
or a dash!
The degree of repetition will depend on the variety of possible symbols.
(a) Morse copde has very few; (b) the English alphabet has 26, (c) a
written code with distinct symbols for whole syllables will have many more,
and (d) a written code with distinct symbols for individual words will have
more still. That is not rocket science.
However, the authors have established a statistical method for determining
which of the above categories (a) - (d) a linguistic text falls into.
Their conclusions are very revealing. The Pictish symbols stones are either
(c) syllable based or (d) word based - depending on whether you accept the
classification in Allen and Anderson in 1903 or the classification done by
Mack in 1997. (Mack classifies the symbols into only 43 categories.)
So, really, the findings are wholly dependent on how you choose to classify
the symbols in the first place - the more you group them together, the more
you can say they reoccur. The conclusions you get for the Pictish symbol
stones alter radically depending on whose classification you use.
This brings us to the other finding - the most important one - that the
repetition in the Pictish symbol stones shows that they must be linguistic
rather than purely pictorial symbols. This finding is based on a
comparison with the higher degree of repetition found in heraldic shields
from Britain in the period 1086-1400.
Now, my first comment here is that it is rather disturbing that only one
data set was used to arrive at a figure for repetition in non-linguistic
symbols. I don't see how that establishes an acceptable 'range'. (Compare
this with the very large number of different types of linguistic texts that
were used to establish ranges for these.)
And second - bear in mind that we are looking at the characteristic of
repetition here, and that will depend on how tightly we lump the symbols
together by reducing the overall 'vareiety'. The authors chose to
categorise the symbols on heraldic shields as follows (p 13): in addition
to noting colour changes, "a simplified set of characters was ... generated
by using only base symbols, e.g. ... all the different lion charges such as
rampant or passant are classified as a 'lion' character". I can't see how
that is a valid methodology when we know for a fact that the variation in
the lion chargees has a significance for their meaning. The authors have
increased the repetion scores for the shileds by grouping together symbols
which are known to be distinct.
So, the scores for the Pictish symbol stones and for the heraldic devices
seem to depend very largely on how one decides to classify the symbols in
the first place. The statistical 'results' simply records what those