These are some measures of the average number of words in a paragraph for a few texts that have been mentioned in this thread.
For the VMS, I used the Zandbergen-Landini transcription, ignoring dubious spaces and using '<$>' as a paragraph marker. The number of paragraphs in Q20 might be slightly underestimated, hence words per paragraph might be closer to 30 than to 36.
I could find online transcriptions of the two Alfonsine works. For the Bologna ms I used the transcription linked by Nick.
I split the BNF ms pointed out by Nick into two parts: the initial lexicon and the main body of text (352 numbered recipes). Between the two there is a third shorter section (a You are not allowed to view links.
Register or
Login to view.) that does not look promising: entries are short and mostly consist of numbers.
The measures for the BNF ms are less reliable, since they are based on extrapolation from my rough counts of a few pages. I guess that the actual measures might differ by +/-10%.
VMS_Q20_recipes Nwords:10092 Npar:280 AvgWordPerPar:36 [30?]
VMS_Pharma Nwords:2168 Npar:44 AvgWordPerPar:49
Alfonso_X_Lapidario Nwords:81227 Npar:322 AvgWordPerPar:252
Alfonso_X_Astromagia Nwords:5261 Npar:251 AvgWordPerPar:21
Bologna_ms_2861 Nwords:46798 Npar:392 AvgWordPerPar:119
BNF6741_lexicon_2r15r Nwords:5000 Npar:200 AvgWordPerPar:25
BNF6741_main_text Nwords:48000 Npar:352 AvgWordPerPar:136
In my opinion, these figures suggest that the two recipes collections (the Bologna ms and the BNF main text) are not good matches for the Pharma section and even less so for Q20: paragraphs are much longer than in the VMS.
The Lapidario is even worse: at least 5 times longer than average VMS paragraphs.
The remaining two texts appear to be closer:
1. Zodiac degrees in Alfonso's astromagia book BAV Reg.Lat.1283a has paragraphs whose length is comparable with Q20. But paragraphs here are highly repetitive: they don't look much like Q20. For instance, these are the figures I get for MATTR 100 using Nablator's java class:
Astromagia:0.55
VMS_Q20:0.84
Looking at the first four degrees in the manuscript (You are not allowed to view links.
Register or
Login to view., Aries is missing) it is easy to see how rigid the structure of these paragraphs is. Nothing like the VMS:
Enel primero grado del signo de tauro sube u(n) omne que trahe un to-ro. El qui nasciere en el sera lazrado & traba-iador & amara mucho el mundo.
Enel segundo grado sube una mugier que tiene enla mano adu fle. El qui nasçiere enel sera amador de solaz & de estrumentos & de todas cosas que pertene-cen a alegria.
En el tercero grado sube una donzella desnu-ya. El qui nasciere en el sera pobre & lazrado &
Enel quarto grado sube una mu ¶ desnuyo. gier que tiene en la mano un estrument quel dizen queber. El qui nasciere en el sera assaca-dor de nueuas & de muchas palabras & grand
¶ mintiroso
2. The lexicon in BNF Lat.6741 (from You are not allowed to view links.
Register or
Login to view. to f15r). You are not allowed to view links.
Register or
Login to view. is like a colour-making-dictionary, with alphabetically sorted entries. While alphabetical order is a problem when matching against Q20, I think that the alternation of blue and red initial letters in these pages is an interesting parallel with the coloured stars in Q20. The number of pages (26) is also comparable with Q20 (24, but with some missing pages).
Obviously, each entry in the BNF lexicon starts with a different word [EDIT: nope]. On the other hand, about 20% of Q20 paragraphs share the same initial word. Also the idea that Q20 might be a lexicon has its problems, but at the moment this seems to me a better candidate than the others. It will be interesting to see if we can find a collection of more-or-less uniformly short recipes.