The evolution of P/F vs T/K raises important questions

[Jorge_Stolfi]

The three short first lines in 3c are f86v3.15, f86v3.18 and f86v3.29, not the titles on f104v, You are not allowed to view links. Register or Login to view. and f108v. Their transcription-defined blocks contain 3, 5 and 8 lines respectively.

Ah, thanks! So this would be an example of puzzling statistics created by improper joining of sections....

Of the 13 first lines shorter than 20 glyphs in 2a, 12 come from f67r2, and the remaining one starts at f99r.47.

I don't think it is wise to count those "mini-parags" in the sectors of f67r2 as parags.  

The nine-puff first line in the Bio group is f76r.1.

Here too, I do not see why that page should be included in Bio.  It has no illustrations, but has those letters in the left margin.  The top half seems to be a single parag with an extra-fancy initial.  It may turn out to be the cover page of Bio, like You are not allowed to view links. Register or Login to view. is presumably the cover page of Herbal; but we should not expect it to have similar statistics.  I would put it in an "Unk76r1" section by itself. 

Same for the text only pages of quire 14, that is, f86v6, f86v5.  They may be a single text, but they may also be independent texts about two completely different topics.

Likewise, the three pages on quire 14 with "Cosmo-like" illustrations (f85r2, f86v4, and f86v3) could be single section, but it may be safer to put each in a separate section until they are clearly shown to be similar.

The current data do indeed count f84v.1–2 as one paragraph.

Like the mini-parags of f67r2, I don't think it is wise to count those two lines as a parag. Ditto for f84v.42-45 and f84v.46-51.

The three very long 4c blocks you identify contain 261, 364 and 293 words, but only 1, 2 and 4 puffs. Their segmentation certainly deserves examination, but they are not all high-puff outliers that inflate a positive correlation.

OK, but they could still distort the (total puff) x (total words) correlation, even in the opposite direction.  For the purpose of this specific analysis, they are a different class of parags than the rest of Stars.  Good that you plotted them separately in the second set of the new plots.

Here is MS section, and by layout. The charts now show fits with free intercepts and actual correlations.

Great, thanks!  Here is what I see in the first set, (head puffs) x (head glyphs):

• Herbal-A and Herbal-B both indicate significant correlation.  In Herbal-B the correlation would be even stronger if the two outliers with fewer than 30 glyphs were excluded.  The fitted line of Herbal-B seem steeper, but that may be an illusion due to the different horiz scale.  Curiously the intercept is negative in both cases, suggesting that puffable glyphs only begin to occur  after the first 10 glyphs or so in Herbal-A, or 25-30 glyphs in Herbal-B.  Or could that be an artifact of the transcription -- namely, the very fancy initial puffs and split gallows seen on many pages are encoded as weirdos that were not counted?
  
• It looks like the correlation in Pharma would disappear if the outliers were excluded.  In Pharma that would be the single head line with ~12 glyphs. That is f99r.47, right?  Unlike most other parags, that one is in a very cramped place that forced a very short head line.  Presumably that tight situation distracted the Scribe away from his puffing habit.   
  
• In Bio too the correlation would probably disappear if all the (non-)parags with short head lines (triangles, with less than 25 glyphs) should be excluded, and the single 9-puff parag too.
  
• In Stars the correlation is low, and would probably be lower in that nine-puff outlier was excluded.  The intercept seems to be close to zero, suggesting that puffable glyphs are uniformly distributed all along the head line.  Maybe because in Stars there are no extra-fancy puffs that would be encoded as weirdos?
  
• Quire 14 (the big fold-out) is a mess.  As I said above, I would just exclude it as being a bunch of single-page  "Unknown" sections.
  

As for the (total puffs)x(total words) plots, they are still suspicious because of possible distortion by omitted parag breaks.  But anyway:

• There seems to be a significant correlation in Herbal B, even if the triangles, the small parags (less than 25 words), and the large outliers (more than 70 words) are excluded.  This suggests significant use of puffs in the body.
  
• In Herbal-A, Bio, and Stars, it looks like the correlation would disappear if the outliers were excluded.  Namely, parags with over 100 words in Herbal-A and Stars, and over 150 words in Bio; as well as the Star parags with 7 or more puffs.  These outliers are probably cases of merged parags.
  
• There is no correlation in Pharma, even without excluding any outliers.  Presumably the reason is that the page layout makes the parag breaks impossible to miss.
  
• Quire 40 again is a mess that is not worth the bother.
  

These observations suggest that, apart from herbal-B, there is little use of puffs outside the head lines.

The third set of plots confirms that there is no correlation of puff use in normal parags of Stars.  The lone starless parag of Stars (third plot on top row) must be the "postface" f116r.31-50, correct?  I would put that whole block in another separate "Unknown" section, since its contents may well be completely different from that of the starred parags.

I can't identify the outlier in the first plot of the bottom row, with ~60 words and 5 puffs.  Is it f102v2.30-36? In my transcription it has ~71 words (counting commas as spaces) but only 4 puffs.

Overall, these results seem to confirm the popular hypothesis that the puffs are fancy glyphs that are randomly substituted for other glyphs or glyph combinations mostly to highlight the head lines of parags, with no other semantic meaning.  Except that occasionally (especially in herbal-B) they are also used to highlight scattered words in the body of parags.

The anomalous margins of You are not allowed to view links. Register or Login to view. and f112v, due to bad vellum surface, as well as the irregular margins of some Herbal pages that are determined by the illustrations, strongly indicate that the line breaks in parag text were chosen by the Scribe based on available space, ignoring the line breaks of the Author's draft.  Therefore, it would seem that the decision to puff up some glyphs in the head lines was made by the Scribe himself, not by the Author.

All the best, --stolfi