Os Lusíadas, 1572: the exemplars and the editions

3.1 Definitions

Here it is fitting to offer some clear definitions which will guide those readers who may be more familiar with textual and literary methodologies, rather than with material bibliography. Edition is a technical expression from the field of bibliography. In hand printing, the type is set by a compositor, line by line, who will then impose the set type within a chase, which is a metal frame holding the type. The editions of Os Lusíadas dated 1572 are in quarto format, which is to say, each sheet has four pages on each side or eight pages in total.^[2] If two sheets are printed from the same forme, without any change in the type, these two sheets are said to be from the same edition. Technically, therefore, edition refers to every copy of every sheet printed from substantially the same setting of type.

In hand printing a number of changes in the forme can occur. The most radical change, and one which does not offer any complication to the definition above is this: the printer completes his work, the forme is returned to the compositor, who then distributes the type back onto the type-cases. In other words, all four pages are taken apart and no longer exist. If the workshop later decided to print more copies of this sheet, they would repeat the entire process of type-setting, imposing and printing. To reset and reprint an entire forme from scratch is to create a new edition.^[3] It may be a new edition of the whole book – every single forme is reset – or it may be a new setting of a specific forme.

Various intermediate changes are, however, possible. For example, someone might notice an error when the printer has already printed one or more sheets. The printing process could be halted, and one would extract and change a specific type, word, line or stanza and then resume printing. This is called an in-press or stop-press correction. If the earlier sheets are not discarded, we would have variant states from the same edition. Variant states may also occur by an accidental movement or loss of type during printing, or by accidental damage to a forme. Unintentional changes may also occur if, for some reason, the printer re-imposes a forme. As there is no limit to how much one can change, the two situations here described – new edition and state – must be seen as a continuum, with potentially a middle area which may be difficult to define precisely. Bibliographers have addressed this, but it need not concern us here at all, for Os Lusíadas is, in this regard, exceedingly simple.^[4]

3.2 Mechanical collation applied to Os Lusíadas, 1572

To answer the main question at hand here: the existing copies of Os Lusíadas dated 1572 constitute two separate editions from a bibliographical point of view. A complete copy of Os Lusíadas has forty-six sheets and two half-sheets, and each sheet is printed on both sides, therefore, the first edition required ninety-two formes for the sheets. Regarding the half-sheets, two formes in total were used (one forme per half-sheet), probably using the technique of work and turn.^[5] Objectively, all the ninety-four formes were set, printed, distributed, re-set, re-printed and distributed again. This, the present study contends, is an objective, bibliographical reality.

Whether António Gonçalves or someone else was responsible for the new edition, whether one is genuine and one pirated, whether both editions were printed in 1572 or not, these are all separate questions. Such questions follow the admittedly intriguing fact that so many mixed copies exist. This article will, later, clarify some specific cases of hybridism, yet the overall problem demands further research to be answered convincingly. For this reason, this author hopes to provide more definitive answers to these related questions in future. As for the fundamental question regarding the number of editions, the present article answers it by stating an objective fact about type-setting: every single type, in every single forme – including lines, words, isolated letters and even the spaces in between the words – was set from scratch twice.

In order to demonstrate this fact, one must resort to mechanical collation, a method of comparing two specimens of the same page. It differs from textual collation in that it is not a comparison of the text taken from two books. Rather, it superimposes the image of two pages and is particularly suitable for comparing the type-setting of printed books. The first optical collator was developed in the 1950 s by Charlton Hinman (1947, 1953, 1963) to address similar problems on Shakespeare’s First Folio. Hinman was able to place two copies of the First Folio simultaneously in his equipment, which allowed him to see exactly where the compositors had introduced specific changes. In other words, the equipment showed where variant states occurred within each page.

Two type-settings of Os Lusíadas exist, and this can be seen in images produced by a digital collator, a piece of software similar to Hinman’s collator, though based on digital technology. Unlike Hinman’s collator, it does not require the actual physical volume for its comparison. Instead, it relies on photographs of early printed books which the software places one on top of the other, and in this way compares the differences. If the printed surface of two exemplars was printed from exactly the same type (same edition), the two images coincide perfectly; if they were printed from two different formes (two editions), they will not coincide at all; if they were printed from a single forme which was corrected or altered in any way (variant states); the images will coincide only where the type has not been changed.^[6]

In selecting exemplars, it was decided that they should be clean, legible and as marginalia-free as possible. However, one is also constrained by copyright permission from certain libraries. The copies thus selected for this demonstration are:

Rio de Janeiro, Academia Brasileira de Letras, shelf-mark LC-C192L;

Oxford, Wadham College Library, shelf-mark A7.24;

Lisbon, Biblioteca Nacional de Portugal, shelf-mark CAM. 1P.

Figure 1

fol. 1^r from Academia Brasileira de Letras, LC-C192L.

Figure 2

fol. 1^r from Wadham College Library, A7.24.

Figure 3

fol. 1^r from Biblioteca Nacional de Portugal, CAM. 1P.

Figure 4

fol. 1^r from Academia Brasileira de Letras and Wadham College Library collated.

Figure 5

fol. 1^r from Academia Brasileira de Letras and Biblioteca Nacional de Portugal collated.

Figures 1, 2 and 3 show the original photographs of fol. 1^r taken from three distinct copies. Figures 4 and 5 were produced by the digital collator. The former superimposes figures 1 and 2, the latter figures 1 and 3. Figure 4 confirms that fol. 1^r from the copies at Academia Brasileira de Letras and Wadham College Library are from the same edition. This figure is presented here in order to show how the collator works by superimposing images that remain legible. One notices only a small blurred area in the printed surface, mostly in the bottom left corner. The reason for this is, undoubtedly, that the photograph of the Wadham College Library copy is not perfectly flat. The binding in this copy is stiff so that the book cannot be fully opened.

Figure 5, in turn, shows the entire printed area extremely blurred. It proves that this whole page was set independently twice. One edition is represented by the Academia Brasileira de Letras copy, the other by CAM. 1P. at Biblioteca Nacional de Portugal. The Academia Brasileira de Letras and Wadham College Library copies correspond to the edition traditionally called Ee; CAM. 1P. at Biblioteca Nacional de Portugal correspond to the edition known as E. With hindsight, Figure 5 allows further observations, listed by printed line (rather than line of poetry):

1. Line 2: de lvis de vs. de lvis de (Roman vs. Italic type);

2. Line 3: camões. (more widely spaced in one than in the other);

3. Line 4: Primeiro vs. primeiro (upper-case vs. lower-case P);

4. Line 9: mares (followed by comma in one but not in the other);

5. Line 17: foram vs. forão (ending);

6. Line 19: Africa vs. Affrica (single vs. double f);

7. Line 19: andaram vs. andarão (ending);

8. Line 19: deuastãdo vs. deuaſtando (short-s plus t ligature and tilde on a vs. long-s plus t ligature and an);

9. Line 22: Cantando (in which the capital C is distinctively different in each);

10. Line 22: espalharey vs. eſpalharey (short-s ligature vs. long-s ligature);

11. Direction line: catch-word Ceſſem (distinctive capital C).

One needs not mention the two most famous cases, widely discussed in previous secondary literature, namely, the difference between Entre vs. E entre (line 14) and edificáram/sublimáram vs. edificarão/sublimarão (lines 14 and 15).

In a second test, folio 185^r from the Academia Brasileira de Letras and the Biblioteca Nacional de Portugal copies were collated. This page is curious because of the last line of the middle stanza, which reads “A naufragios, a pexes, ao profnndo.” It has a typographical error in the last word, which should read profundo (‘deep’), rather than profnndo, which is meaningless in Portuguese. The

Figure 6

fol. 185^r from Academia Brasileira de Letras, LC-C192L.

Figure 7

fol. 185^r from Biblioteca Nacional de Portugal, CAM. 1P.

Figure 8

fol. 185^r from Academia Brasileira de Letras and Biblioteca Nacional de Portugal collated.

compositor apparently confused the type n for u, two similar letters. Agudo (1972) observed that all copies dated 1572 have this typographical error. For this reason, he rejected the possibility of two editions, on the basis that two compositors would not concur in this same error. The collation of this page, however, confirms that everything, including the line in question and the meaningless word, were set independently twice.

Once again, figures 6 and 7 show the original photographs, while figure 8 is the superimposed image of these two, produced by the digital collator. As with figure 5, in figure 8 the printed area became extremely blurred.

The main reason for the disparity between the two images is the spacing. In hand printing, compositors would have at their disposal a variety of different spaces, that is, blank pieces of type. In later years, from the widest to the narrowest, they were Quadrats, Em spaces, En spaces, Thick spaces, Middle spaces, Thin spaces, Hair spaces. Early printers had a less-clearly defined, but probably equally diverse, palate of spacing material to choose from. The wider the space, the easier it is to distinguish, whether by eye or touch. But as the spaces become narrower, it becomes increasingly difficult to tell them apart. Given the amount of type being handled by a professional compositor and the speed of work, it was almost impossible to choose consistent spacing from the various inter-word spaces available. Indeed, the case from which the compositor took the type probably held all these different-width spaces within a single compartment.^[7] By setting spaces of different widths between the words, the two lines will not overlap perfectly in mechanical collation, even if they look the same, and read identically, to the human eye.

In order to confirm this hypothesis, it is possible to zoom into the lines with the typographical error. Figure 9 shows the last line of the second stanza taken from the Academia Brasileira de Letras copy at the top and Biblioteca Nacional de Portugal CAM. 1P. at the bottom. The two images were aligned and dotted lines were added to draw attention to significant differences in spacing. The comparison shows that in the first example the compositor did not use any space after commas, whereas in the second the compositor did. Furthermore, the distance between “ao” and “profnndo” is wider in the latter.

Figure 9

line 16 from fol. 185^r from Academia Brasileira de Letras (top) and Biblioteca Nacional de Portugal (bottom).

Once again, these two copies represent different editions in the strictest sense. The two compositors committed the same typographical error twice or, alternatively, a first compositor incurred this error and a second sought to copy the original slavishly, to the point of reproducing even this kind of meaningless error.

4.1 A method for distinguishing the editions: stanza initials

The simplest, most convenient method relies on a comparison of certain initials. By initials, it is meant here only those that open each stanza, for instance, the E which opens the second stanza in Figures 1, 2 and 3; the E, O, P, each opening one stanza, in Figures 6 and 7. The large A in Figures 1, 2 and 3 cannot be used, neither can the equivalent large initials which open each Canto. The initials at the beginning of each line of poetry are not particularly helpful either. As a rule, each page has three initials which can be relevant.

In the Academia Brasileira de Letras and Wadham College Library copies, the first initial of each stanza tends to be an Italic type. In Biblioteca Nacional de Portugal CAM. 1P. these same initials are predominantly in a Roman typeface. Folio 2^r offers a fine example of this principle, where the Academia Brasileira de Letras copy displays three italic initials, E V V (Figure 10), whereas Biblioteca Nacional de Portugal CAM. 1P. displays three roman initials, E V V (Figure 11).

Figure 10

fol. 2^r from Academia Brasileira de Letras, LC-C192L.

Figure 11

fol. 2^r from Biblioteca Nacional de Portugal, CAM. 1P.

The observation that one edition uses predominantly Italic initials while the other uses a large number of Roman initials is not in itself original. Rather, it is well documented in the secondary literature. However, the present study proposes that this feature must be systematically adopted henceforth when discussing the 1572 editions of Os Lusíadas.

It is useful to outline here why this particular feature is optimal. Firstly, it does not require access to the actual copies. Anyone with a facsimile or photograph can distinguish Italic from Roman initials. An expert can be guided by other features: the paper texture, chain-lines, wire-lines and watermarks. Even the smell of certain copies can help identification. However, these features are not only harder to transmit but are also only available in the originals.

Secondly, edition and new edition are defined as printing the text from different settings of type. It is exclusively a typographical problem. The initials are, precisely, a typographical reality. This is not to say that the paper used in each edition is not relevant; on the contrary, the stocks of paper used in one and the other are different. Yet by focusing on initials, one achieves a very tight system based on one definition and one objective feature.

Thirdly, the initials affect almost the entire copy. Other systems, such as Ee versus E, or S versus D, fail on this account. These identification systems suggest that every copy with the left pelican and the E entre reading are alike and belong to one edition, and that all copies with the right pelican and the Entre reading are also alike and belong to the other edition. However, as Jackson (2003) correctly observed, British Library G.11286 mixes these features: a left pelican on the title-page and the Entre reading on fol. 1^r.

This article proposes therefore that these earlier designations should be abandoned. The pelican relates only to the title page while the designation Ee versus E relates just to fol. 1^r. The initials, instead, run throughout almost the whole book. Thus, Ee or S may be more appropriately called “Edition with Italic initials”; E or D may be called “Edition with Roman initials”. More simply, one can call them Italic edition and Roman edition.^[9]

In order to guide readers, this article offers a catalogue of formes of the two editions of Os Lusíadas which can be used to identify which sheet belongs to which edition (Table 1). Each sheet is identified by its first signed page, and the sheet is subdivided in outer and inner forme, that is, to the two sides of the sheet. The initials of the opening stanzas are then listed under two columns, one for the Italic and one for the Roman formes. There are usually three stanzas per page and

Figure 12

The first sheet of Os Lusíadas.

hence three initials, except when Cantos start and end. For this reason, fols. 1^r and 19^r only feature one initial each.

For those less familiar with material bibliography, a few words might be useful here. If we were to dismantle a copy of either the Italic or Roman edition, back in the sixteenth-century, we would be left with a number of sheets of approximately 36.5 cm×26 cm (Fig. 12).

Because the book is a quarto, the sheet is folded twice, giving four folios (or leaves) or eight pages. The folios pictured above are, left to right and top to bottom: fol. 7^r, fol. 2^v, fol. 8^v and fol. 1^r. If we were now to reassemble our copy, we would fold two sheets in half, lay one on top, prick and stitch the centre folding (Fig. 13).

Figure 13

Two folded sheets.

This constitutes our first gathering or quire, composed of two sheets, being therefore, a quarto in eights. To help printers and binders, some pages have signatures on the bottom, which indicates the sequence of text. Our first image represents a sheet signed A1.A2.A7.A8; on the second image, the sheet in the foreground would be A3.A4.A5.A6, with the centre fold and stitching occurring between A4 and A5. The next gathering would be signed B, with the same numerical sequence, and so on, until gathering Y. Gathering Z is exceptional in that it has ten folios. When bound, the leaves of the book were still connected on top, and then readers had to cut that in order to read the book. For this reason, if we were to dismantle a copy today, we would not have a full sheet, but rather pairs of leaves (bifolia) connected where the book is stitched.

Returning to the catalogue of formes, it seemed redundant to list initials for every single page; as the book is a quarto, one out of four pages is enough to identify each forme. Those seeking to analyse copies of Os Lusíadas with this catalogue, should however ensure that the leaves analysed are conjugate. Conjugate means that two leaves (for example A1 and A8) are still connected, that is, they constitute a bifolium, as one would expect. When damage occurs, two leaves may become disconnected.

Each gathering is made of two sheets, which requires four formes. The following example should make it clear how the analysis must be conducted:

A1.A8; A2.A7; A3.A6; A4.A5 — these pairs must be conjugate (a bifolium);

A1.A2.A7.A8; A3.A4.A5.A6 — these are the two sheets within gathering A;

A1^r.A2^v.A7^r.A8^v — first forme;

A1^v.A2^r.A7^v.A8^r — second forme;

A3^r.A4^v.A5^r.A6^v — third forme;

A3^v.A4^r.A5^v.A6^r — fourth forme.

These formulae are valid for most gatherings, from A to Y. The exception is the last gathering, Z:

Z1.Z10; Z2.Z9; Z3.Z8; Z4.Z7; Z5.Z6 — these pairs must be conjugate;

Z1.Z2.Z9.Z10; Z3.Z4.Z7.Z8 — these are the two sheets within gathering Z;

Z5.Z6 — this is a half-sheet;

Z1^r.Z2^v.Z9^r.Z10^v — first forme;

Z1^v.Z2^r.Z9^v.Z10^r — second forme;

Z3^r.Z4^v.Z7^r.Z8^v — third forme;

Z3^v.Z4^r.Z7^v.Z8^r — fourth forme;

Z5^r.Z6^v — fifth forme;

Z5^v.Z6^r — sixth forme.

The first two leaves – featuring the title-page, alvará and censura – are not signed and constitute an independent gathering. It is conventional to refer to it by the Greek letter π. Its two leaves, π1 and π2, must be conjugate.

The catalogue reveals that whereas both editions mix Roman and Italic type, the Roman edition is the most irregular of the two in this regard. There are indeed some pages which cannot be distinguished with this system, when the Roman edition employs three italic initials (fols. 9^r, 91^r, 137^v, 177^r, 179^v, 181^v). Nevertheless, the general rule highlighted here holds true for most of the book. In those pages which cannot be distinguished by initials, one can usually compare another page of the same forme, for instance:

B1^v.B2^r.B7^v.B8^r — fol. 9^r with Italic initials; use fols. 10^v, 15^r or 16^v.

S1^v.S2^r.S7^v.S8^r — fols. 137^v and 138^r with Italic initials; use fols. 143^v or 144^r.

Z1^r.Z2^v.Z9^r.Z10^v — fols. 177^r and 186^v with Italic initials; use fols. 178^v or 185^r.

Z3^v.Z4^r.Z7^v.Z8^r — fol. 179^v with Italic initials; use fols. 180^r, 183^v or 184^r.

Z5^v.Z6^r — fol. 181^v with Italic initials; use fol. 182^r.

Unfortunately, in one forme, M3^r.M4^v.M5^r.M6^v, every initial in the Roman edition is Italic (fols. 91^r, 92^v, 93^r, 94^v), meaning that, in this particular case, the two settings cannot be distinguished by initials. The catalogue offers, instead, two distinctive lines from each setting, with differences highlighted in bold type.

The first half-sheet, gathering π, does not have any poetry and, therefore, cannot be distinguished by initials either. Instead, the catalogue provides some distinctive features from the two title pages (π1 recto) and from the royal alvará (π2 recto). These are amply recorded in previous secondary literature. The three formes in question – M3^r.M4^v.M5^r.M6^v and the two π formes – are the only cases in which a system based strictly on initials fails to be functional.

4.2 Alternative methods for distinguishing editions

As summarized earlier in this article, the academic community has moved from a relative consensus that there are two editions of Os Lusíadas dated 1572 (even if other matters were polemic) to a more intense debate on whether there are two or just one edition. Why does such debate occur? The view adopted here is that such discussion hinges more on matters of terminology than empirical observation. There has been little or no explicit discussion in previous studies regarding what each scholar understands by edition. Hence this study started by offering a formal definition of edition. This is compounded by the fact that E/Ee or D/S are themselves inadequate terminologies.

Recently, Alves and Tocco have called for additional investigation focusing on the books’ physical characteristics, showing that they were not convinced by Ruas’s earlier work, which discussed paper watermarks.^[10] The different stocks of paper used for the editions have already been mentioned, but it is worth addressing this question: can they, or other material aspects, help in determining the number of editions?^[11] Unfortunately, the answer is not simple.

What this particular scientific community really needs is, first and foremost, an unanimous terminology. Without such prerequisite, material investigations cannot add much to the debate. If the scientific community were to use the expression edition in its strictest, technical sense, the problem becomes much simpler. It becomes a question merely of whether there are one or two independent typesettings, in which case mechanical collation offers the ultimate proof. Edition, as a technical expression relating to typesetting, is unaffected by the paper stocks used.

That said, this author is documenting the paper of every sheet of every copy of Os Lusíadas dated 1572, which are then classified in categories, or paper types. Paper types provide useful evidence, but they should be used a posteriori, rather than a priori. That is to say: we define two editions as two independent type-settings. Given that this occurs, we need a simple system to identify them, which is given by contrasting Italic and Roman initials.^[12] Once a given sheet specimen is identified as belonging to either of the two editions, then one can analyse the paper of this particular sheet.

Hypothetically speaking, even if such investigation were to reveal that every copy of the Roman and the Italic edition was printed in the exact same paper, we would still have two editions. Yet the analysis undertaken for this study has confirmed Ruas’s conclusions: the two editions were printed in different stocks of paper. This strongly suggests that only one edition is genuine and the other is most likely a fraud. If António Gonçalves was setting both the Roman and the Italic edition in 1572, the stocks of paper should be less independent. Given that they are totally independent, one must admit other hypotheses: another printer copying the first edition in 1572, António Gonçalves making a second edition at a later year, another printer active at a later year. These would all constitute fraud, given that both the Italic and the Roman claim to have been printed by António Gonçalves in 1572.

For reasons of space, this section will offer a very brief explanation of what the analysis of paper has so far revealed. A more detailed explanation is reserved to a later study. Hand-made paper found in books of this period was made in pairs of moulds, also called twins, and the analysis of paper thus normally reveals such pairs. In the Italic edition, the author has confidently identified 32 paper types:

1. eleven pairs of watermarked paper;

2. three types of watermarked paper (their respective pairs not yet found);

3. three pairs of unwatermaked paper;

4. one type of unwatermaked paper (its respective pair not yet found).

In the Roman edition, there appear to be fifty types of watermarked paper and ten types of unwatermarked paper.^[13]

No single copy of the Italic edition contains all the 32 paper types, but rather a fraction of them. An average of sixteen paper types per copy is a good estimate. Likewise, no copy of the Roman edition contains all its fifty paper types. It is therefore more appropriate to refer to two separate “groups” or “universes” of papers types. By comparing a large number of copies, one may eventually catalogue these two universes. Every Italic sheet known to this author is always printed on one of the paper types belonging to the first universe (noting that Italic sheet is defined by looking at its typesetting, not its paper type). Every Roman sheet so far analysed, equally, is always printed on a paper type belonging to the second universe. The author has never found a paper type which has been used for both the Italic and the Roman editions. This means that, if comparing a pure Roman and a pure Italic copy, there is no overlap whatsoever between the paper types. If comparing two pure Italic copies, there will be many similarities and some differences regarding paper types, but all paper types belong to the same universe. The author has not yet seen two pure Roman copies. The next section will explain in more detail the phenomenon of hybrid copies, for there are indeed copies which alternate Roman and Italic sheets. Obviously, such copies present us with samples from both universes of paper types. Yet the overall observation remains valid: every Italic sheet within these copies was printed on paper types from the first universe, whilst the Roman sheets were printed on paper types from the second universe. Thus, in short, the analysis of paper types confirms the independence of the Italic and Roman editions.

4.3 A tripartite method based on formes

A system of identification based on initials is not the only possible system but, as shown here, it is by far the simplest and most convenient. Alternatively, one could devise a system based on specific readings from each forme, though with no advantage. It is worth, however, spelling out why the scientific community needs a system based on the analysis of formes.

Once scholars can identify and talk about formes as units within a book, many difficulties which we now currently face are resolved. Thus, for instance, the proposed terminology can operate on three levels.

Firstly, one can speak of an Italic and a Roman edition. This expression refers to an unknown number of sheets printed in the sixteenth century. This is an example of how the expression can be used: in Noronha’s hypothesis, the Italic edition was printed by António Gonçalves in 1572 whereas the Roman edition is a later counterfeit. The meaning of this sentence is: every sheet which we identify today based on Italic initials (but which are distinctive in a number of ways) was printed by António Gonçalves in 1572; every sheet which we identify based on Roman initials was printed by someone else at a later date. Some of these sheets survive today, dispersed in various copies, some were lost over the centuries. Each edition is the entirety of the sheets, both extant and lost sheets. In summary: scholars can speak of Italic and Roman editions, yet the printing of these two editions is an historical phenomenon to which we have only indirect access.

Secondly, one can designate a specific sheet as Italic, Roman or neither. For example: in Academia Brasileira de Letras, LC-C192L the π sheet (based on the title-page) and the first A sheet (based on fol. 1^r) are from the Italic edition; yet in British Library G.11286, the π sheet is from the Italic edition, whereas the first A sheet is from the Roman edition.

Thirdly, one can and must distinguish the actual copies. For example: in Academia Brasileira de Letras, LC-C192L all sheets come from the Italic edition, making it a pure Italic copy.

Unlike edition, to which we have only indirect access to, the living scholar has direct access to sheets and copies. The copy is a useful, self-contained unit; the sheet is a smaller unit within the copy, for those cases when we need to discuss its constituent parts in more details. Smaller units are available, but, as a rule, the sheet suffices.

An alternative unit, smaller than the sheet is, of course, the forme; and the catalogue of formes offered in this study breaks down each sheet into two formes. However, having analysed twenty-five copies in loco, as well as some facsimiles, this author has never detected a single case of a sheet which is printed from the Roman edition on one side and from the Italic edition on the other side. It is almost certain that such a sheet never existed for, in all likelihood, the two editions were printed at different periods. Thus, while it is good practice to distinguish each forme, in reality, we do not require such a level of detail: one forme per sheet is enough. Incidentally, this also addresses the problem of forme M3^r.M4^v.M5^r.M6^v: its other side, M3^v.M4^r.M5^v.M6^r, is distinctive enough on the basis of initials and will probably suffice for anyone analysing the internal M sheet.

Another example of a unit which is smaller than the sheet is the leaf. As a rule, each group of four leaves belongs to the same sheet, based on the formulae given above. This rule is only broken when the leaves are not conjugate, in which case discussing leaves rather than sheets is more appropriate. This happens on some copies of Os Lusíadas dated 1572, but it is a rare phenomenon. Therefore, for the vast majority of copies, referring to sheets is adequate, but, in some exceptional cases, one must refer to specific leaves.

4.4 The six categories of copies

Only the analysis of units – sheets, formes and leaves – allows one to classify copies. The existing copies dated 1572 appear to fall into four main categories, with the last one subdividing into three sub-categories:

1. Pure Italic: a copy in which every sheet belongs to the Italic edition.

2. Pure Roman: a copy in which every sheet belongs to the Roman edition.

3. Incomplete: a copy which lacks one or more folios which have not been replaced by any printed folio.

4. Mixed copies: copies with sheets or leaves from more than one edition. Subdivided into:

5. Hybrid: a copy which has sheets from both Italic and Roman editions. In this case one may specify whether the copy is predominantly Italic or Roman.

6. Interpolated: a copy which has at least one printed sheet which does not belong to either Italic nor Roman editions.

7. Composite: a copy in which one or more folios are not conjugate and it can be demonstrated that these folios originate from different sheets.^[14]

The first two categories are self-evident, while the other categories invite further clarifications. Thus, this article will now demonstrate how and why the twenty-five copies analysed in loco fit into these categories. Conceivably, the analysis of more copies in future could expand the number of categories, yet the four main categories proposed here are unlikely to change.

The copies so far analysed are:

Brazil, Rio de Janeiro:

Academia Brasileira de Letras, LC-C192L

Biblioteca Nacional, C,29,29A

Biblioteca Nacional, C,29,29 (ex.2)

Instituto Histórico e Geográfico Brasileiro, ARM.OR

Real Gabinete Português de Leitura, no shelf-mark

France, Paris:

Bibliothèque nationale de France, RES P-YG-38

Bibliothèque nationale de France, RES-YG-74

Italy, Naples:

Biblioteca Nazionale di Napoli, S.Q.24.G31

Portugal,

Coimbra:

Universidade de Coimbra, Biblioteca Geral da Universidade, Cofre 2

Guimarães:

Sociedade Martins Sarmento, no shelf-mark

Lisbon:

Academia das Ciências de Lisboa, BACL Cofre 1-27

Biblioteca Nacional de Portugal, CAM. 1P.

Biblioteca Nacional de Portugal, CAM. 2P.

Biblioteca Nacional de Portugal, CAM. 3P.

Biblioteca Nacional de Portugal, CAM. 4P.

Biblioteca Nacional de Portugal, CAM. 11P.

Oporto:

Ateneu Comercial do Porto, no shelf-mark

Spain, Madrid:

Biblioteca Nacional de España, R-14207

Biblioteca Nacional de España, R-14208

UK,London:

British Library, C.30.E.34

British Library, G.11285

British Library, G.11286

Oxford:

Bodleian Library, Antiq.e.P.1572/1

Wadham College Library, A7.24

Private #1 (copy in a private collection, whose owner prefers to remain anonymous).