The distribution of Eva-f and Eva-p

[ReneZ]

There was apparently a typo in the table in the opening post, with an almost negligible impact on the result. Anyway, I redid the counts, also for a number of interesting parts of the MS.

First of all:

- All text

- All text marked as A language by Currier

- All text marked as B language by Currier

   


This shows that almost all properties are virtually the same:
- 19% (one fifth) of all words in the first lines of paragraphs have at least one p.
- there is only 0.5 to 1% of such words in the rest of the paragraph text
- there is a visible difference in the ratio p / f: 3.0 for A language and 3.7 for B.
This last point is caused by the number of words with p. There is no difference in the fraction of words with f.

The same can be done for quires 13 (biological) and 20 (stars / recipes) which are both B language:
   

The main thing to be observed is that in these two parts the ratio p / f is clearly increased, to over 5 for both.

However, and in summary, the distribution of f and p is remarkably consistent throughout the MS, even in texts that have otherwise quite different properties.

One more thing that can be seen from these statistics is the average number of lines per paragraph. These can be estimated from the nr of word tokens in the first and all other lines. Using the terminology in the tables:

(P-1 + P-n) / P-1

This gives:

All text:   5.2
A lang:    5.9
B lang:    4.9
Q 13:      9.0
Q 20:      3.6

PS: a number of pages have not been assigned an A or B language and these are only included in the 'All text' statistics.