Stop being analitical!

[Torsten]

Torsten, you're making two different points and it's important to state them explicitly:
1) that the difference between the occurrence of k and t is due to the space available and the space each glyph takes; and
2) that there is a relationship between the frequencies of similar words.

On 1), this is the point I was addressing in my last comment. Your observation on their occurrences at the start of lines is not sufficient to support your hypothesis that this is due to their shapes. It is a supposition which is contradicted by a simple observation: both k and t increase in frequency when o is added and t only declines once q is also added. Your hypothesis would need t to become less frequent once no longer at the start of a word.

On 2), I agree that there are relationships in the frequencies of similar words. But different sets of related words have different relative levels. There is not one single pattern, but multiple patterns. These patterns are seemingly determined by the glyphs present in a word. They are not "bumps and slacks" but important aspects of what's actually happening. They are important because they can disprove several theories of how the text was created.

1) Yes, both k and t increase in frequency when o is added. But at the same time d decreases in frequency when o is added. This happens since o + d is normally transformed into o + gallow. Therefore you have to sum up all the -dar, -kar and -tar word types to see the real picture:

*ar (413)  o*ar (294)  qo*ar (226)
*al (296)  o*al (294)  qo*al (257)
*an ( 24)  o*an ( 12)  qo*an ( 11)
[* stands for d, k, t]

You can also sum up all the -dor, -kor and -tor words:

*or (122)  o*or ( 88)  qo*or ( 67)
*ol (202)  o*ol (190)  qo*ol (152)
*on (  0)  o*on (  0)  qo*on (  0)
[* stands for d, k, t]

2) Indeed, different sets of related words have different relative levels. The text is changing over time from Currier A to Currier B. Therefore is is obvious that it is not a single pattern, but multiple patterns. This is also expected since a text from the medieval ages must be the result of an human mind. A plain cone without any bumps and slacks would prove that the text is the result of an algorithm and was probably generated with a device like a computer.

Your objection is that the statistics must be plain and simple. But there is no need to follow a given set of rules without any exception. Not only that it would be very hard to do so manually, the text generation method would also be far to obvious this way. Being creative is a characteristic of the human mind.