Bifolia created using previous as reference

[Jorge_Stolfi]

Using my recovered spatial coordinates, i can gather two measurements: margin offset, and text angle.

Wait, there seems to be some problem with your data. Here is the clip of your first "black" image for the bifolio f1+f8:
     
Ho do you measure those coordinates? The vertical coordinate seems to be from the first text line on the page.  Is the horizontal one measured from the edge of the Beinecke Library (BL) image?

The plot for You are not allowed to view links. Register or Login to view. seems OK, except that the top margin is omitted.  But note that the width (~15 mm) and slant (practically vertical) of the left margin were determined by the binding, which may not have been where the fold was while the bifolios were still unbound (if they were folded at all).  Moreover, the width of that margin is somewhat affected by the bending of the vellum around the rest of the quire; which you cannot have corrected for, because there is no text there.

The plot for You are not allowed to view links. Register or Login to view. seems wrong. Besides ignoring the fact that the text starts halfway down the page, the actual left margin seems to be only a bit wider than that of You are not allowed to view links. Register or Login to view. (~19mm) if measured from the actual vellum edge, and  fairly uniform (within ~1 mm), except that it seems to follow the curvature of the edge of the vellum.  

The plot for You are not allowed to view links. Register or Login to view. seems a bit off too.  It too ignores the extra-broad top margin (~80 mm).  On the BL images, the width of the left margin (next to the gutter) seems to be more uniform than shown in your plot, varying from ~17 mm at the top to ~15 mm at the bottom.  Which is about the same as that of f1r.  The left rail (the left boundary of the text area) also seems to be more vertical than what is shown on your plot.  However, both measurements are complicated because of the evident bending of the vellum (see the top edge on the BL image) and because there seems to be a sharp and deep crease just to the right of the binding fold.  My guess is that this crease is where the bifolio was originally folded, before the binding. The right text rail is also fairly parallel to the right edge of the vellum, except that in the middle part the latter bulges a bit to the right, while the rail bends slightly to the left, then to the right for the last three lines -- so that the width of the right margin, top to bottom, is ~27 mm, ~28 mm, and ~26 mm. 

The plot for You are not allowed to view links. Register or Login to view. is rather misleading too.  The left rail of the first parag is indeed tilted CCW relative to the (free) vellum edge, but only by ~3 degrees, whereas the plot has ~7 degrees. That rail is quite straight and the width of the left margin is pretty uniform increasing only  from ~22 to ~23 mm from top to bottom; except that the left edge of the vellum (but not the rail) is irregular and shifts to the left by ~2 mm near the bottom of that parag.  The left rail of the second parag is not collinear with the rail of he first parag (like the plot suggests).  Instead, the rail there is tilted ~5 degrees relative to the vellum edge and starts at~24 mm from it, several mm to the left of where the rail of the first parag would be.  As for the right margin, it starts at ~28 mm but quickly decreases to ~25 mm by line 5, and then remains quite uniform until the end of the second parag, at ~24 mm.

While these errors would not matter when considering each page in isolation, they seem to matter a lot when comparing margin widths across bifolios.

For example, You are not allowed to view links. Register or Login to view. has significant slant.

Do you really mean f20v?  Compared to the spine, its baselines are quite horizontal, and the left rail is tilted only ~1.5 degrees CCW.  The free left edge of the vellum is tilted ~1.5 degrees in the other direction, so the angle between the two is only ~3 degrees -- which seems within the normal range of variation for our Scribe. And the right rail seems to be quite vertical, except that the lines do not quite end on it.

When arranged in my proposed golf ordeering, these horizontal offsets and offset angles form clear sequences. I am not claiming this is the perfect order. But it is clear that it is close, and many pairs likely do capture true production sequence. We can see that the angles and offsets from previous bifolio are frequently copied to the next.

But you are claiming that the Scribe used the previous bifolio as a guide to define the margins of the next one.  Why on Earth would he do that?  The margins vary all over the place, even because the edge of the vellum is often irregular and tilted.  Nowhere we see evidence that he cared about that sort of consistency.  Check the Zodiac diagrams, which any professional draftsman would have made all of the same size, with equally spaced nymphs etc. 

Aren't you falling again for the "Book Sorting" fallacy?  Namely, if you sort a bunch of things by similarity of an attribute X, you will find an apparent gradual evolution in the value of X -- even if (or especially when) the X of each thing is an independent random variable.

All the best, --stolfi