Construction of the manuscript

To the main page

The beginning and end of the book

The first page of the book looks exactly like the next 57 pages. However, this is quite normal in a manuscript; elaborate title pages were not common before the introduction of printing.

The last page is blank and contains scribblings. This is also quite normal - the scribblings are usually pen trials with little relevance to the contents. This implies that the last quire was intended to be the last, and it may conceivably be an index or table of contents. The last paragraph, offset from the rest of its page and lacking the characteristic stars, may conceivably be a colophon (a statement by the scribe with a description of the contents and a date).

The structure of the paragraphs

John Grove has pointed out f 67rb as the only place in the manuscript where the text is written to fit ruled lines. It is noticeable that the body of the letters does not sit on the line as in Roman script, but that the ascenders of the gallows characters hang from the upper line and the descenders of characters like y sweep down to the lower line.

In other parts of the manuscript it is possible to identify notional or invisible rulings (perhaps pencilled and erased) from which the ascenders hang and on which the descenders rest. I have manually added such rulings to images of the stars section: of course there is a danger of divining imaginary 'ley lines' by joining up a random set of points, but I am impressed how often ascenders and descenders align even in the densely written B section. I have pointed out before that the script in sections such as the biological and stars sections is abnormally cramped. As I now see it, a line of text occupies three zones - ascenders, body and descenders - and wherever possible ascenders and descenders rest on the invisible rulings. However, since space is lacking, the same space has to be used for the lower zone of one line and the upper zone of the next. At some points the body of a line wavers downwards to avoid overwriting the descenders of the line above, but its descenders do not, so far as is possible, waver with the body; they simply get truncated so as to meet the invisible ruling at the expected point. At the end of a line, ascenders rise to touch their ruling even if they were forced down earlier in a line

Occasionally (there are examples on f 104r) an ascender sweeps down beyond the nearest ruling to rest on the next one further down (i.e. the descender occupies two zones not one). Such long descenders seldom get overwritten by the next line as if the writer knew in advance that the space would be free, and to explain this I have previously suggested that the text was written in alternate lines. I have never been certain of this, but I think that there is a problem to be solved here.

Within a paragraph the rulings are usually parallel to the extent that this is possible, but the groups of rulings in successive paragraphs are often out of mutual alignment. Did the writer use a pencil and ruler and slip from time to time?

The structure of the pages

Upper margins

The descenders of the first line of a page do not hang from an imaginary line. If anything, they are aligned with the irregular edge of the vellum sheet. The first paragraph of 105r begins lower down the page than expected, but that is because the column of line-initial letters begins lower down than it need do.

Where text is divided into two columns by part of an illustration, as on 75r and 75v, the upper margins of the two columns do not align exactly, but the fact that the two columns have the same number of lines indicates that the text is a unit, probably meant to be read across the division. The obvious explanation is that an original text was written across the divide but that the copyist first wrote out all of the left hand column and then the right hand one.

Left and right margins

The right hand margins of pages are not closely aligned: it seems that the copyist preferred to create a ragged right hand edge rather than split a word between two lines.

The left hand margins of right hand (recto) pages have a strong vertical alignment, as if the first action of the copyist was to create an imaginary vertical division of the whole blank sheet in front of him. The left hand margins of left hand pages tend to be shakier, as if the scribe was using the boundary of the vellum as a guide. However, some of the left hand margins display a firm vertical alignment, particularly in the stars section.

At certain points, individual lines and paragraphs stop short of the right hand margin of a page. The lines of 78r do not continue to the margin and then run over: they look very like lines of poetry. The first paragraphs of 112r and 112v are very intriguing. The first 22 lines of 112r are about 20% shorter than the others: they share the left hand alignment of the whole page but stop short of the right hand margin in a ragged alignment. Overleaf, the first 22 lines of 112v are also shorter than the rest. However, they do not align with the left hand margin of the whole page but conform to their own vertical alignments (two of them) indented rightwards. The implication is that the relevant corner of the sheet is a copy of an original which was physically smaller than the surrounding paragraphs.

At other points, it seems that groups of paragraphs are not written to fit left and right hand margins. For instance, paragraphs 2 to 12 of 111r constitute a distinct block of text: within the block, the start of a paragraph is marked only by a marginal star and all the lines extend to the right hand margin of the page, as if the copyist knew that he had too little space to start each paragraph on its own separate line: yet paragraphs 13 to 17 are spaced normally. There is no very obvious reason to copy different paragraphs in a different way, and the distinction may very well reflect a feature of an original manuscript which was being copied.

In all sections of the text, whatever the layout of the right hand margins, the final words of lines display the same statistical peculiarities (a preponderance of two and three letter words, final m, final ry and ly). But where it seems that two paragraphs have been run together on one line, there is a reduced frequency of gallows characters at the beginnings of lines (however, other characteristics of line-initial words, such as a high frequency of initial d, s and y, are not affected).

There is no general alignment of paragraphs across the gutter of a quire and the text area of facing pages is roughly but not exactly the same size. However, the illustrations on 78v and 81r are two halves of one drawing: the pipes connect with each other over the gutter. This confirms the suggestion of Currier that the quires were copied from loose leaf, unbound originals.

Lower margins

The lower margins of pages raise no particular puzzles except for the occasional case in which the last line of a page (e.g. 105r) is aligned with the central axis of the page and not the left margin. This is difficult to explain.

Preliminary explanation of observations

As Prescott Currier observed years ago, numerous features of the appearance of the manuscript indicate that it is a copy and not an original. Further more, what I have said above (particularly about 75rv and 112rv) indicate that the copyist aimed to reproduce the layout of the original and largely succeeded at the expense of a certain amount of cramping. It looks in fact as if each line of the manuscript corresponds to a line of the lost original.

It was not normal to copy a manuscript page by page and line by line, even where the original cried out for such treatment. For instance, the huge biblical commentaries called Postillae consisted of windows of text surrounded by columns of glosses and commentary, much like the Talmud. Incredibly, the complex calculations required to align text and commentary were performed anew at each copying: the page divisions of different copies are not the same. To take another example, university stationers sometimes hired out manuscripts quire by quire for copying - but the resultant copies ('pecia' manuscripts) are characterised by marginal notes indicating where one quire of the original ended and another began, and these marks are generally in the middle of a page.

I suggest that the Voynich copyist took such pains because the original text was itself enciphered. It is known (again from the work of Currier) that the first and last words of a line have distinct statistical properties which appear to be an artefact of the encipherment scheme: suposing that this was known to the copyist, he had reason to cram his letters into the space provided. Why then does the copy look cramped? Some possibilities suggest themselves.