University of Virginia Library

The first editions of Don Quixote, Part I (1605), Novelas ejemplares (1613), Don Quixote, Part II (1615), and Persiles y Sigismunda (1617) were printed at the Madrid press of the widow of Pedro Madrigal, which in those years was under the direction of Juan de la Cuesta.[1] Of these four Cuesta assignments, the setting and printing of the first edition of Novelas ejemplares was in most respects the smoother and more straight-forward job. For example, with the exception of only eight pages (out of a total of five

                                           

Type	Occurrences
B	K3, 4; O5v, 22; S7v, 1; T1v, 26; V1v, 18; X6v, 16; Y5v, 18; Bb4v, 21; Cc4, 7; Ee2, 18; Ff8v, 27; Gglv, 34; Hh7, 23
C	A3, 15; B2v, 8; H1, 13; I7, 6; K3, 25
C	A6, 20; F1v, 17; G3, 15; O2v, 19; R8, 18; T5v, 17; V8v, 9; Y7, 6; Z8, 8; Aa1, 1; Bb7v, 14; Cc5, signature; Gg4v, 30; Hh6, 5; Kk6, 13; Ll4, 14; Mm1, 21
C	D4v, 8; E7v, 34; G5, 31; N5, 12; X6, 27; Z6, 3; Bb5v, 23; Cc6v, 2; Dd6v, 10; Ee3v, 28; Ff6, 28
C	Ff1, 21; Hh8v, 31; Mm1v, 30
D	Bb6v, 21; Ee5, 27
E	Y2, 3; Z8, 19; Aa2v, 8; Dd6v, 23; Ee4, signature; Ff5v, 20; Gg3, 31; Hh5v, 17; Ii3v, 16
E	B7, 27; D5v, 20; M8, 14; N4v, 32; O5v, 27; T8, 6; Z8v, 9; Aa7v, 4; Ff8v, 6
E	B7v, 16; C5v, 14; D3, 20; O4v, 24; P5, 9; Q4, 27; S1v, 13; Y3v, 7; Z8, 5; Aa2v, 4; Bb7, 22
E	Ll5v, 22; ¶¶2, 27
G	Q1v, 6; V1v, 4; Kk1v, 12
H	B1, 25; D4v, 15; E7v, 31; F7v, 21; G8, 12; H5, 22
M	A3, 33; G5v, 10
M	G6v, 22; K8, 13; L6v, 28; M2, signature; O3, 28; T6v, 32; X1v, 9; Dd2v, 9; Gg4, 29; Ll4v, 34
M	Ff2v, 14; Gg8, 7; Hh2, 34; Ii8, 12; Ll8v, 32; ¶¶7v, 17
N	A7v, 5; D6v, 23; H6, 19; I7v, 9; L4, 18; M6, 33; N8v, 16; P1, 32; Q8, 22; R2v, 27
O	A8v, 33; B7v, 24
Large P	C1, 26; L2, 16; ¶3, 26
T	P3v, 31; Z1, 18; Ff6v, 8; Hh6, 28
V	A6v, 25; B1v, 1; C6v, 16
Y	T5, 9; V6, 4; X4v, 24; Y4, 24; Aa2v, 11; Cc7, 1

hundred and seventy-one pages—page ¶lv is blank) all other pages of this work were composed from the same set of type cases; see Table 1.[2]

Table 1 contains only a small sample of the dozens of types which could be used to show that only one set of type cases was used throughout most of the job (I shall return to the eight irregular pages later on), but I have listed only those types which are easy to spot and identify and which together interrelate all the gatherings with each other.

Another telling characteristic is the relatively small number of running titles used in this edition. In the following table I have entered the running titles per forme and entered the formes according to the order in which they were usually completed. This order, however, does not necessarily obey the sequence in which the formes were printed; but on the whole the order of printing followed the order in which the formes were completed. Pages A1, I2, L7, Q6, V6v, Dd4, and Gg1 have no running title.

Table 2 shows that only seventeen running titles were used, and that of these running titles only four do not appear throughout the book: running title a (first appears in gathering E), running title b (appears last in gathering Q), running title i (appears last in gathering F), and running title q (first appears in gathering D). I did not include gatherings ¶, ¶¶, and Mm in the table because gatherings ¶ and ¶¶ have no running titles, and because gathering Mm has only four pages: pages Mm1 and Mm2v made up the outer forme, and pages Mm1v and Mm2 the inner forme (page Mm1 has running title a, page Mm1v has running title d, page Mm2 has running title m, and page Mm2v has no running title).

With the help of Table 2 we can follow in some degree the way in which the job advanced. It is evident that even though some running titles appear in consecutive formes (running title d, for instance, appears on signatures A2v and B3v—last forme of gathering A and first forme of gathering B), or fail to appear for a period of time (running title m appears on signature E4, does not appear in gatherings F or G, and reappears on signature H8), the more common sequences have from one to four formes in between occurrences of the same running title (running title m, signatures A7 and B3; running title i, signatures B2 and C3; running title n, signatures A7v and B7v; running title g, signatures A4 and B5). The periodicity in which the running titles alternate from the beginning to the end of the alphabetical sequence of gatherings clearly shows that the printing of this work was for the most part uneventful. The periodic reappearances of any given running title are, of course, neither predictable nor regular, but it is evident from Table 2 that the setting of type, and the imposing and running off of the formes was a steady process devoid of most of the complex setting and printing patterns one finds in other works printed at the Madrigal-Cuesta press.

A third helpful aid is the striking orthographic consistency one is faced with throughout the first edition of Novelas ejemplares. Some examples: I found eighty-eight occurrences of the spelling variant vuessa merced and thirty-seven abbreviations (v.m., thirty-four occurrences; vs.ms. three occurrences), but I found no occurrence of the spelling variant vuestra merced, which was the other common spelling used for this form of address.[3] Moreover, the thirty-seven abbreviations were clearly meant to stand for the form vuessa merced, because whenever the word vuestra appears abbreviated it appears set as vrtildea. (vrtildea., thirteen occurrences; vrtildeas., five occurrences; vrtildeo., twelve occurrences; vrtildeos., six occurrences), never as v. alone. The only exceptions to this thoroughly consistent rule are the abbreviations used for the forms of address vuestra Excelencia, vuestra Magestad, and vuestra Señoria which appear abbreviated with upper-case V's (V. Excelencia, one occurrence; V. Magestad, eight occurrences; and V. Señoria, three occurrences). One could argue that these upper-case V's might stand for the form vuessa or Vuessa, but the text is perfectly clear in this respect. In my article "The Need for a Scholarly, Modernized Edition" I suggested that in all probability Cervantes used the form vuestra in these and other related forms of address, except when he was making sport of one of his characters; the text of Novelas ejemplares bears this theory out. Whenever any of the forms of address just mentioned above does not appear abbreviated it invariably

occurs in conjunction with the form vuestra, with lower-case v (vuestra Excelencia, five occurrences; vuestra Magestad, six occurrences; and, too, vuestra Alteza, one occurrence; vuestra Christiandad, one occurrence). And I found no instance of an improper spelling of any of these forms of address because at no time are any of them used in a sarcastic or humorous vein, as some of them are on occasions used in Don Quixote.[4]

Two other examples of this orthographic consistency: there are four hundred occurrences of the word volver (to return) and related forms, of which three hundred and ninety-seven occurrences follow the spelling boluer and only three the spelling voluer (which in all probability were caused by the compositors' using foul-case types). There are three hundred and eleven occurrences of the word misma (same) and related forms, all of which follow the spelling misma, vis-à-vis mesma; and yet we know that Cervantes used both spellings because they appear in his other works side by side.

This remarkable orthographic consistency and the fact that only one set of type cases was used throughout most of the job strongly suggest that the entire setting of the first edition of Novelas ejemplares was assigned to one compositor. One of the more reliable methods for finding out whether or not this theory is sound is to obtain the overall quadmark of the work.

The following table shows the numbers of quads set before punctuation marks per page in this work.[5] There might be, and probably are, some minor differences between the numbers noted in Table 3 and the actual numbers of quads per page because at times I found it very difficult to tell with all certitude whether or not there is a space between a word and the following punctuation mark, and because all occurrences had to be collated by hand. But these minor discrepancies would not modify in any substantial way the general conclusions I shall draw from this typographical evidence. When in the following table only one figure is entered under any given page it indicates the number of quads set before commas in that page (see, for example, page A6v; two quads set before commas). When two figures are given, the

second one indicates the number of quads set before other punctuation marks rather than comma (page A1; one quad set before a comma and three quads set before other punctuation marks).

When we add up the full pages of type that have the same number of quads set before punctuation marks and project the various totals over a horizontal guide line ("Quads per page") we obtain the overall compositorial quadmark, which in this instance is a double-peak curve (Diagram 1).[6]

I have argued elsewhere that a double-peak quadmark is likely to result if we have insufficient data, if we consider the work of two different compositors together, if we mix pages set seriatim and pages set from cast-off copy together, or if we have any combinations of these conditions (see "The Compositors", pp. 35-36).

The first possibility should be discarded outright. The typographical evidence we have in this instance is both sufficient and conclusive. The possibility that more than one Cuesta man might have set copy for the first

edition of Novelas ejemplares cannot, on the other hand, be rejected without first entering into an in-depth study of the text.

I have already pointed out that typographical (one set of cases, a relatively small number of running titles) and textual (orthographic consistency) evidence strongly suggest that the entire setting of this edition was assigned to one compositor, but, of course, this theory does not necessarily imply that the job was set by one workman alone. Compositors could either work on their own, in teams but independently from one another (compositors C and D, and compositors E and F; Don Quixote, Part I), in a closer team relationship (compositor I and apprentice Z; Don Quixote, Part II), or in more complex patterns (Don Quixote, Parts I and II). Hence, and in order to ascertain exactly how many men worked on the setting of the first edition of Novelas ejemplares, I shall now delve into the text of this work of Cervantes's.

Despite the striking orthographic consistency one finds throughout the pages of this book, this uniformity is by no means an out-and-out phenomenon. One can find scores of spelling variants in Novelas ejemplares, just as I have found them in the first editions of Cervantes's other works, but, unlike in my previous monographs, I cannot base my study of orthographic variants on frequently used words (vuesa/vuestra, bolver/volver, mesma/misma) because either these words occur in only one spelling or the occurrences of a differing spelling are so few that the different readings might be merely the result of the compositor's setting foul-case types accidentally. I have, therefore, relied solely on words that show few occurrences, keeping in mind the following premises: 1) many of the variants could be authorial and might have been entered deliberately by Cervantes for stylistic purposes,[7] 2) this sort of variants cannot be always trusted because the fewer occurrences one has of any given variant the less reliable one's conclusions might result,[8] and 3) compositors were not always one hundred per cent consistent in their spellings when they set unfrequently used words and often changed from one spelling to another in the midst of a job.[9]

In spite of these obvious drawbacks, I was able to find twelve words that can be used to demonstrate that the first edition of Novelas ejemplares was set by more than one workman. (In Table 4, pages that appear within blocks have fourteen or more quads set before punctuation marks per page).

The parting of the twelve spelling variants listed in Table 4 into two clearly defined groups according to whether the pages where they occur have thirteen or fewer or fourteen or more quads set before punctuation marks per page proves that the first edition of Novelas ejemplares was set by two workmen. Moreover, because most of the preferred spellings of these Cuesta men were so similar, and because both men worked in the same gatherings together and used the same cases of type and the same running titles we can safely conclude that they were a team made up of a compositor and his apprentice. It is now obvious that the Cuesta hand whose habit it was to set quads before punctuation marks whether or not they were needed preferred the spellings agradecelle, agradecida, agradecimiento, alfange, accidente, banco, bulto, continuo, efecto, encaxar, huesso, monja, Precetor, and Tiniente; whereas his co-worker preferred the spellings agradezelle, agradezida, agradezimiento, alfanje, acidente, vanco, vulto, contino, efeto, encajar, hueso, Monja, preceptor, and Teniente.[10] But the latter, who was probably the instructor, was not as consistent in his use of these spellings (see especially variants 1, 10, and 12) as he was with the spellings of more frequently used words, a fact that would explain why the apprentice was able to introduce his own preferred spellings (or retain those of the printer's copy) without being brought to account for it.

The next step is to identify these workmen by comparing some of their preferred spellings of frequently used words to those of the compositors who set the first edition of Don Quixote. And here, what had been a stumbling block when trying to distinguish between the compositor and his apprentice, i.e., the striking similarities between most of their spelling preferences, suddenly becomes an invaluable asset.

In Table 5, I have listed some readings, most of which appear in the stints of compositors C and D (first edition of Don Quixote, Part I), E (first and second Madrid editions of Don Quixote, Part I), F (first and second Madrid editions of Don Quixote, Part I, and first edition of Don Quixote, Part II), and H, I (with apprentice Z), and J (first edition of Don Quixote,

Part II), and in the first edition of Novelas ejemplares. I have indicated with x's the preferred spellings of the various compositors. The numbers of occurrences that we find in Novelas ejemplares appear in the last column under a question mark. When I could not be certain about which variant form was the preferred spelling of a given compositor I left the appropriate spaces blank. For instance, compositor J set only one occurrence of the form apenas (Ee3, 30) and one occurrence of the form a penas (Ll1, 3); hence neither occurrence was entered in the table. Most departures from the preferred spellings used in Novelas ejemplares (see, for example, variant 6, last column, three occurrences of the form cauallero) are either foul-case variants (variant 6: Andres cauallero, E5, 4; Cauallero is a proper name and should have been capitalized, and the compositors were aware of this fact—nine previous occurrences were capitalized: B5v, 34; B8, 17; B8v, 9; C2v, 3; C3, 6; C7, 17; D4, 5; D8, 17; and E2v, 24), authorial variants (variant 6: Cervantes uses the words señor cauallero—with which Rinconete addresses Cortadillo, I2v, 17—and esse cauallero—with which Cortadillo refers to the scoundrel Monipodio, I7v, 28—ironically; see above, footnote 4), or compositorial variants (see the variant encaxar/encajar; Table 4 [variant 8] and Table 5 [variant 17]).

Variants in capitalization do not include occurrences of homographs unless the homographs themselves have variants. For example: in variant 6, Cauallero/cauallero (gentleman) I have not included occurrences of the word cauallero (mounted), but in variant 22, Sol(es)/ſol(es) (sun) I have included the homograph Soles/soles (eyes) because both words have variants.

When we tabulate the overall dominant spellings listed in the last column of Table 5 (variant 6, Cauallero; variant 16, efecto; variant 17, encaxar; etc.) with the dominant spellings of the compositors of Don Quixote we find that the compositors of the first edition of Novelas ejemplares disagree with the preferred spellings of:

• compositor C in variants: 1, 3, 4, 6, 7, 12, 13, 15, 16, 19, 21, 22, 23, and 27.
• compositor D in variants: 1, 2, 5, 6, 10, 11, 16, 19, 21, 22, 23, and 27.
• compositor E in variants: 1, 2, 3, 6, 7, 8, 9, 10, 11, 12, 13, 14, 21, 22, and 24.
• compositor F in variants: 3, 4, 5, 6, 7, 11, 16, 19, 20, 21, 22, 23, 24, and 26.
• compositor H in variants: 1, 5, 11, 15, 16, 19, 21, 22, and 25.
• compositor I in variants: 1, 5, 7, 14, 19, 21, 23, 24, and 26.
• compositor J in variants: 1, 13, 14, 15, 18, 19, 20, 21, 24, and 26.

It is therefore evident that neither of the compositors of Novelas ejemplares was one of the Cuesta men who worked on the first editions of Don Quixote. The differences between the preferred spellings of the compositors are too clear to cause any uneasiness, especially because the majority of the words listed in Table 5 were selected only because they appear in the stints of all or of most of the compositors. I could list dozens of other words which show orthographic differences between only two or three compositors at a time (see variant 9), but I would be merely stressing what is already obvious. I therefore shall from now on refer to the Cuesta hand who set substantial numbers of quads before punctuation marks in the pages of Novelas ejemplares as apprentice Y, and to his instructor as compositor K.

We can now confidently state that compositor K's normal range of quads set before punctuation marks was from five to eleven quads per page, with a peak at between seven and nine occurrences per page (see Diagram 2). And after examining some of the pages set by compositor K we can conclude that he usually set tight full lines of type with narrow spaces between words and punctuation marks ("Gitanif | mo,y la mas hermofa,y difcreta, que pudiera hallarfe,no | entre", A1, 16-18), that he set very few quads per page, and that he used them mainly to justify his lines and normally placed them between words and after punctuation marks ("bufcô | por todas las vias que pudo, y no faltò poeta que fe los | dieffe", A1v, 19-21).

The normal range of quads set before punctuation marks of apprentice Y was from fourteen to thirty-six quads per page, with a peak at between nineteen and twenty-three occurrences per page, but his overall range goes from zero quads set before punctuation marks (pages containing poems or sections of poems; compositors did not usually set unnecessary quads when setting poetry) to fifty-four or more quads per page (see Diagram 2). It was the habit of apprentice Y to spread-out normal lines of type by leaving wide blank spaces between words and before and after punctuation marks. He frequently set unnecessary quads when setting prose ("f;intieſ- | ſe : mas quando boluí de mi deſmayo , y me vi ſolo | en", F4v, 20-22).

It seems to me unlikely that compositor K might have set many pages

with more than sixteen quads before punctuation marks, at least in this job.[11] On the other hand I am almost certain that apprentice Y set a substantial number of pages with fewer than fourteen quads before punctuation marks, but I hesitate to separate the different stints, mainly because they do not always coincide with the printed pages as we now have them; for example, page F4v has thirty-seven quads set before punctuation marks, and the first nineteen lines of page F5 have twenty-one quads (probably all set by apprentice Y), but the last fifteen lines of page F5 have only three quads, and page F5v has only six quads (compositor K?). Nevertheless, I find quadmarks a very reliable tool for verifying overall compositorial patterns because setting, or not setting, quads before punctuation marks was a mechanical compositorial habit which had nothing to do with the characteristics of the printer's copy of works in prose.

On the basis of the typographical evidence provided by the marked differences in the numbers of quads set before punctuation marks per page and of the textual evidence provided by the spelling variants listed in Table 4 we can also begin assigning to one or to the other workman some readings that show few occurrences. For example, the readings vagage (one occurrence, page P1, 5; thirty-six quads), ropage (one occurrence, page Kk1, 29; sixteen quads), barco (one occurrence, page V7v, 21; twenty-two quads), and

raxa (one occurrence, page G5, 14; fourteen quads) agree with the preferred spellings of apprentice Y for the readings alfange, banco, and encaxar (Table 4, variants 2, 4, and 8). And a similar pattern emerges from the spellings preferred by compositor K: the readings vagaje (one occurrence, page D8, 27; eight quads), varco (one occurrence, page N1v, 21; thirteen quads), and raja (two occurrences, pages S4, 7, and Y5, 25; five and zero quads respectively), agree with the related spellings set by compositor K, alfanje, vanco, and encajar (Table 4, variants 2, 4, and 8). Furthermore, we are now able to pinpoint some compositorial variants even when the readings occur on pages which could not be assigned to one or to the other Cuesta man solely on the basis of the numbers of quads set before punctuation marks. The word satanás (Satan), for instance, appears in Novelas ejemplares with two different spellings: Satanas (B4v, 33; and Ll6v, 19) and Sathanas (L5v, 19; P5, 18; and V1, 32). Because all five occurrences of this word appear on pages each of which has less than fourteen quads set before punctuation marks (B4v, seven quads; L5v, thirteen quads; P5, eight quads; V1, thirteen quads; and Ll6v, nine quads) we could have mistakenly assumed that all occurrences had been set by compositor K, and, thus, that his preferred spelling was the form Sathanas simply because there are three occurrences of this form and only two of the form Satanas. But now that we know that apprentice Y had some strong spelling preferences of his own we can safely conclude: 1) that the different forms of the word satanás are not the result of the compositors' changing spellings at midstream, 2) that the spelling difference between the two forms (-t-, -th-) could not have arisen accidentally, 3) that the preferred spelling of compositor K was Sathanas whilst the preferred spelling of apprentice Y was Satanas, and 4) that apprentice Y did indeed set some pages with less than fourteen quads before punctuation marks.

Once we are certain that the workmen who set the first edition of Novelas ejemplares can be told apart and identified through some of their spelling preferences and the number of quads they set before punctuation marks per page, we can begin putting other pieces of this typographical puzzle into place. Now I shall address myself to the questions of whether or not some pages, or gatherings, might have been set from cast-off copy, and, if so, whether or not this fact would modify the quadmarks of compositor K and apprentice Y.

When manuscript copy of a work in prose is cast off compositors almost always receive slightly less or slightly more text than they would otherwise set in their stints when setting seriatim. On such occasions compositors frequently have to depart from some of their usual setting practices in order to make the text of the cast-off copy fit their stints, and must resort to some common tricks of the trade. According to whether they find themselves with less copy than necessary or too much text, the compositors would set either looser or tighter lines of type than usual, impose shorter or longer pages than the average page of the work, set additional or avoid unnecessary rules and quads, and/or develope or use contractions and abbreviations.

Pages of type with an irregular width or length should therefore be

examined closely, especially when they occur in groups. Such is the case, for example, with pages C3 to C6v. These pages belong to, and exactly make up the two formes of the inner sheet of gathering C, and they are four millimetres wider (102 mm) than the average page of the first edition of Novelas ejemplares. Also, page C4 is the only page in the whole book which has more lines of type (thirty-five lines) than the average page (thirty-four lines). These typographical characteristics cannot be regarded as mere coincidences. They strongly suggest that the pages of the inner sheet of gathering C were set from cast-off copy.

What in all probability happened is that compositor K must have found himself lagging behind the pressmen when he was working on gathering C, and must have decided to complete the two formes of the outer sheet first. That way he had to set only eight pages (C1 to C2v and C7 to C8v) rather than the twelve pages (C1 to C6v) he would have had to set to complete two formes set seriatim. Thus, when he finished setting page C2v he apparently cast off copy for pages C3 to C6v, put it aside, and began setting page C7. Once he had finished setting the pages of the outer sheet either he or his apprentice imposed them and gave the formes to the pressmen. Then compositor K returned to the cast-off copy and began setting the pages of the inner sheet. When he finished setting the text he found himself with some twelve extra lines of type, and, because the formes of the outer sheet either had already been run off or were in the process of being printed, he had to absorb the extra lines in the pages he had just finished. It is self-evident that the compositor absorbed the extra lines by increasing the length of his lines of type by four millimetres each and making an irregular page of thirty-five lines of type.[12] Another typographical characteristic that seems to support this theory is the substantial numbers of contractions set in these pages (three hundred and eighty-nine contractions altogether), but to explain this anomaly I must refer to gathering E, where most of the recourses resorted to by compositors when they found themselves with extra lines of type were used (see Table 6).

Despite the considerable differences between the numbers of quads set per page in gathering E (from two quads—page E2—to twenty-eight—page E4v), we must consider all the pages of this gathering regular in this respect because the numbers of quads are well within the extended ranges of compositor K and apprentice Y. These variations, therefore, have very little to do, if any, with whether the workmen were setting seriatim or from cast-off copy.

The irregularities between the numbers of contractions used from page to page are so marked (from zero to forty-two) that they must necessarily reflect some variable conditions of one sort or another.[13] These variations are especially puzzling because they occur throughout the book and in a very irregular sequence where isolated pages, or groups of pages, with few contractions per page alternate with pages or groups of pages with many contractions (see Table 7).

At first I thought that these variations in the numbers of contractions per page revealed another compositorial trademark, that one Cuesta man did not like to use contractions whereas his companion used them frequently. But what in fact seems to have happened is that the main type case used to set the first edition of Novelas ejemplares had only sufficient type to set about nine or ten full pages of text without becoming short of some of the

sorts most often used (a's, e's, o's, and n's).[14] Thus, the compositor (probably compositor K) was able to set pages A1 to A7 (thirteen pages, but the majority of these pages are not full pages of type; they contain the beginning of La Gitanilla—one half page of type—and three long poems) without worrying, but when he was setting page A7v he began to be short of n's (fifteen of the nineteen contractions set in this page stand for n's) even though type from the first forme ran off (A3v, A4, A5v, and A6) had apparently already been washed and was being distributed. As soon as the clean type was added to his case the compositor immediately returned to his usual practice (lines 30 to 33 of page A7v have no contractions; page A8 has only three contractions.)[15]

The running off of the formes and the washing and distribution of type kept pace with compositor K until he reached the middle of page B8 (lines 1 to 22 have only four contractions; lines 23 to 31 have seventeen).[16] (It is important to notice that pages A8 to B7v make a total of sixteen pages, exactly four formes of type.) As compositor K was finishing page B8, however, he must have received clean type from a previous forme (lines 32 to 34 of page B8 do not have any contractions) and was thus able to set the equivalent of four full pages of type (pages B8v, C1, C1v, and C2—pages C1 to C2 are not full pages because they contain two short poems—and the first half of page C2v—lines 1 to 15 have only eight contractions) without having to set more contractions than was his usual custom. But he realized then that some sorts were becoming scarce and began setting more contractions than usual (lines 16 to 34 of page C2v have twenty contractions). As long as distribution

of type and composition kept pace with one another (pages A1 to C2) the compositor was replenishing his case with type coming from formes which had very few contractions (for example, the outer forme of the outer sheet of gathering A—A1, A2v, A7, and A8v—has only five contractions), but whenever composing out-paced distribution the compositor was forced to use more contractions per page (forme) and hence he began receiving clean type from formes whose numbers of contractions varied greatly from one to the other (for instance, the outer forme of the outer sheet of gathering C— C1, C2v, C7, and C8v—has one hundred contractions; the inner forme of the inner sheet of gathering C—C3v, C4, C5v, and C6—has two hundred and seven contractions, the outer forme of the outer sheet of gathering D—D1, D2v, D7, and D8v—has only twenty-seven contractions). Consequently, one would look in vain for the regular pattern of full pages of type with few contractions based on the number four (number of pages per forme) or multiples of four (two formes = eight pages; etc.) to continue beyond the first few gatherings. Nor is there any relationship between numbers of quads set before punctuation marks and numbers of contractions set per page.

There is, on the other hand, a close relationship between the numbers of contractions and abbreviations set per page. We have already seen that compositor K and apprentice Y set relatively few abbreviations of the form of address vuesa merced. If we collate the pages where most of these abbreviations occur with the pages that contain substantial numbers of contractions we will find out, not surprisingly, that of the forty-four pages with abbreviations thirty-five have more than ten contractions per page.[17]

All this typographical evidence makes it clear that the striking differences between the numbers of contractions and abbreviations set from page to page are not the trademarks of two workmen, nor the result of setting from cast-off copy.[18] These irregularities reflect instead the varying states of the

case as the job progressed. Whenever compositor K and apprentice Y set abbreviations and substantial numbers of contractions they did so only because they were short of type.

With these conclusions at hand we can explain away the overlappings shown in Table 6 ("Contractions", and "Quads set before punctuation marks") and the differences in the numbers of contractions set in the pages of gathering C, and can at last return to explaining the anomalous width and length of pages E4 to E5v (Table 6). What seems to have happened at this stage of the job, as in the case of gathering C, is that compositor K had to hurry up with the setting of gathering E (probably because the pressmen were idle and he did not have a complete forme ready for them), whereupon he cast off copy for pages E4 to E5v and gave it to his helper as soon as apprentice Y had finished setting type for page E3v. Then compositor K began setting page E6 (see the numbers of quads set before punctuation marks in these pages: E3(19), E3v(12), E4(11), E4v(28), E5(22), E5v(22), all set by apprentice Y; and E6(2), E6v(2), E7(7), E7v(7), E8(4), E8v(7), all set by compositor K). By doing so compositor K saved some precious time because instead of one compositor having to set five pages to complete one forme (apprentice Y would have had to set pages E4, E4v, E5, E5v, and E6 to complete the first forme of the gathering—pages E3v, E4, E5v, and E6) apprentice Y and compositor K had to set only six pages between the two of them (E4 to E6v) to complete two formes (inner and outer formes of the inner sheet; see Table 2), and keep the pressmen busy. If both workmen were setting simultaneously they could not, of course, set from the same case, as indeed they did not. One has only to compare the upper-case C's, G's, and P's used in pages E4 to E5v with those used elsewhere in the book to realize that these four pages were set from a different case (cf., for instance, the upper-case types that appear on pages E3v, lines 26, 28, 31, 32, and 33; E6, lines 2, 3, 4, and 7; and E6v, line 8, with the upper-case types that appear on pages E4, lines 2, 5, 7, 10, and 11; and E5v, lines 24, 28, and 32).

When apprentice Y finished setting his stint he found himself about eight lines short of type, and thus began shortening the length of his lines by three millimetres each, but this reduction was not enough and he ended by having to impose these pages one line shorter than the average page.

There must have been some rush to give work to the pressmen, because as soon as apprentice Y had finished fixing page E5 either he or, more probably, compositor K broke the normal order for imposing and running off the formes and imposed the outer forme of the inner sheet (E3, E4v, E5, E6v) before the inner forme, which contained the last page of the stint of apprentice Y (E3v, E4, E5v, E6), and gave it to the pressmen to run off. We know that this is what in all probability happened because there is an error in a running title that supports such a theory. Running title h appears complete on page B8v, it appears without the middle "e" of ex[]mplares on pages C3v and E6v, but it appears fixed on page E5v (and elsewhere). It is obvious that the outer forme of the inner sheet must have been printed before the inner forme of the inner sheet. And thus we come to understand the reasons that lie behind all the anomalies (shorter and narrower pages of type, and different typefaces) and seeming irregularities (numbers of quads set before punctuation marks, and numbers of contractions used per page) that we find in gathering E of the first edition of Novelas ejemplares. Some thought, however, should be given here to the identity of the Cuesta man who set pages E4 to E5v. Could he have been a third compositor setting from his own case of type rather than apprentice Y setting from a case usually assigned to another Cuesta man? I think not, for two reasons: two occurrences of one of the preferred spellings of apprentice Y appear precisely on one of these pages (see Table 4, variant 12), and the numbers of quads set before punctuation marks (eleven, twenty-eight, twenty-two, twenty-two) are well within the range of apprentice Y.

There are a few other irregular pages in the first edition of Novelas ejemplares (see footnote 6. Also, page S8 has an irregular width of one hundred millimetres), but these anomalies are in all probability the result of minor adjustments or miscounting. The casting off of copy when setting prose works from manuscript copy makes more sense when two or more compositors are setting simultaneously and are using different cases of type (as during the setting of the first editions of Parts I and II of Don Quixote, and gathering E of the first edition of Novelas ejemplares), but accomplishes relatively little when only one compositor is at work or when the compositors are both setting alternately and using the same cases of type, as compositor K and apprentice Y obviously were throughout most of the job. Therefore, I would conclude that even though these Cuesta men might have at times absorbed a few extra lines of type by using contractions and abbreviations, and might have on occasions expanded their lines of type by setting extra quads, their favourite recourses were either to increase (pages C3 to C6v) or decrease (pages E4 to E5v) the length of their lines of type according to the particular situation in which they found themselves. But these conclusions acquire weight only after one has understood, explained, and demonstrated that there were two compositors at work in the setting of the first edition of Novelas ejemplares, that each compositor had his own characteristic quadmark, and that the variations in the numbers of contractions and abbreviations used per page reflect the state of the case at any given moment. Were

we asked to provide further textual evidence to uphold these theories we would not have to go too far afield in our search for this evidence.

No two copies of the first edition of Novelas ejemplares are identical, of course, but most typographical differences reflect only the natural wear and tear of type or minor accidents that occurred during the running off of the formes. Some copies of this book, however, differ from one another in other respects. Copy C.59.b.20, The British Library, has four pages (Ee4, Ee4v, Ee5, and Ee5v) which were not printed from the same formes used for printing those of copy G.10181, The British Library. The running titles that appear on pages Ee4 to Ee5v of copy G.10181 belong to the set of running titles used throughout the book (running titles k, n, k, and l; see Table 2), thus we can confidently assert that the four pages of this copy are part of the original full inner sheet of gathering Ee. Moreover, because copies G.10181 and C.59.b.20 differ from one another only on these four pages, because two of the four running titles used in the pages of copy C.59.b.20 belong also to the original set of running titles (running titles m, Ee4; and p, Ee5v), and because some of the types used in the running titles show more wearing on pages Ee4 to Ee5v of copy C.59.b.20 than in the following pages where these running titles appear, we can conclude that both copies belong to the same edition, that pages Ee4 to Ee5v were set twice to make up a short run, and that the pages of copy C.59.b.20 belong to the second printing of the sheet.[19] I will distinguish between these two copies following the same system and nomenclature I used when dealing with copies of the first edition of Don Quixote, Part I (see The Compositors, pp. 33 and ff.). Copy G.10181, The British Library, belongs to family group A of the first edition of Novelas ejemplares; copy C.59.b.20, The British Library, belongs to family group B. Furthermore, because the compositor who re-set pages Ee4 to Ee5v accidentally retained an obviously incorrect reading that occurs in the original pages (socorrido—in lieu of the correct feminine form socorrida—copy G.10181, Ee5, line 32; copy C.59.b.20, Ee5, line 33) we can ascertain that he used the printed rather than the authorial manuscript pages as his copy. But these assertions are all foregone conclusions. More important would be to answer the questions: why did the compositors have to re-set these four pages? And when and by whom were they re-set?

One telling characteristic of pages Ee4, Ee4v, Ee5, and Ee5v is that they occur on perfected sides of the same half of the inner sheet of gathering Ee. Therefore the second printing of these pages cannot be the result of a last-minute shortage of copies of the inner sheet of the gathering, otherwise the

compositors would have had to re-set the pages of both formes. What apparently happened is that someone must have spilled something over, or somehow damaged this half section of a considerable number of sheets.[20] Hence the accident must have occurred when the sheets were stacked together waiting to be folded. The facts that some types of the running titles used in the second printing of pages Ee4 to Ee5v show more wear in the re-set pages than in the last gatherings of the book, that the running titles used in pages Ee4 and Ee5 (second printing) came in all probability from the very last page of the book with text (Mm2) and from the last forme of gathering Ll, and, finally, that the compositor had to set two new running titles (probably because most of the original running titles either had already been distributed or were still in press) strongly suggest that the resetting of pages Ee4 to Ee5v probably took place when compositor K was finishing, or had just finished setting the last gathering of the book.

The workman who re-set the pages made a considerable number of changes, but the changes introduced are all incidental, cancel each other out, and/or have very few occurrences; hence, they do not help us to pinpoint the identity of the compositor who made them.[21] I have shown elsewhere, however, that the Cuesta compositors changed some spellings of their copy as a matter of course. So, it is most surprising that only four spelling changes were made in these pages, especially because the changes might not be, after all, spelling variants but rather the result of the compositor's wanting to save space (por que to porque, and si no to sino; we must keep in mind that the compositor expanded one hundred and ten contractions, and, consequently, had to set his lines of type tighter to allow for these changes) and a foul case (Yglesia to Iglesia; there is an occurrence of the form Yglesia just four lines below the "spelling variant", suggesting that the change might not have been intentional). It logically follows from these premises that the compositor of the re-set pages must have been either compositor K or his apprentice, who are unlikely to have introduced spelling variants in the text they had already regularized. Had the compositor set these pages from the same case of type used throughout most of the job we would have

been able to reaffirm this contention, but we would have found ourselves on a dead-end road because we would not have been able to ascertain precisely which of the two men set the pages belonging to the second printing. Fortunately, the compositor of the re-set pages did not use this case, he resorted instead to the case used earlier to set pages E4 to E5v of this edition (cf. the upper-case types used in the two sets of pages), which also happens to have been the case from which compositor K took the long upper-case Y's he needed when he was setting page N2v (see above, footnote 15; the only upper-case Y's used in the re-set pages appear on page Ee5, lines 14 and 15). Now then, had compositor K been the person who re-set pages Ee4 to Ee5v he would have almost certainly used his own case of type. Therefore we must conclude that the Cuesta man who re-set these four pages was apprentice Y. Furthermore, because apprentice Y did not use the case of his instructor we can surmise that both men were probably setting simultaneously (as they did when setting gathering E), and that compositor K must have been working on the last pages of the book. Hence, the more likely sequence in which running title m might have been used is: inner forme of the inner sheet of gathering Ll (page Ll6; see Table 2), re-set pages (page Ee4), and page Mm2. What could have happened, and in all likelihood did happen is that when apprentice Y finished re-setting pages Ee4 to Ee5v compositor K was still at work on pages Mm1v and Mm2. So he probably imposed pages Mm1 and Mm2v (colophon only!) together with his pages Ee4 and Ee5v (twin-half page imposition), and thus the pressmen had a forme to run right away. Then apprentice Y set two new running titles and had pages Ee4v and Ee5 ready and waiting to be imposed with the last pages of the book. Once both sides of the sheets had been printed and the ink dried, the workmen folded the sheets and cut them in half.

The pressmen did not have to run as many copies of the re-set pages as of pages Mm1 and Mm2v because they needed only enough copies of pages Ee4 and Ee5v to make up for the number of sheets which had been damaged. Consequently, they must have taken the re-set pages out sometime during the run of this forme. And thus the type of pages Ee4 and Ee5v could have been distributed before compositor K had imposed the last forme of gathering Mm. This sequence of events would explain how running title m could have been used in both formes (pages Ee4 and Mm2). The whole process could have been slightly different, but this hypothetical reconstruction of what might have happened seems to me the most plausible and practical way of dealing with this problem.

From the changes that apprentice Y made to his copy we now can double check and polish some of the conclusions reached earlier. Because he expanded one hundred and ten contractions and developed two abbreviations in spite of the fact that he had to fit (and did fit) his text into exactly four average pages proves that he did not like to contract or abbreviate and, thus, that the numerous contractions and abbreviations he set in the first edition of Novelas ejemplares are not the result of a compositorial habit nor of his need to absorb extra text. This dislike for setting abbreviations and contractions

is especially evident because of all the hundreds of changes he made, only five are contractions, and because compositor Y set these five contractions on the very last lines of the last page of the stint (Ee5v, second printing, lines 22, 25, and 33: fatisfacion>fatiſfaciõ, que>qtilde, acompañaſſedes >acõpañaſſedes, cuenta>cuetildeta, and que>qtilde). it is obvious that he set these abbreviations only when he found himself forced to fit the remaining thirteen and one half lines of his copy text in the thirteen lines he had left if he was to end with a regular thirty-four line page. This fact proves that apprentice Y used contractions (and probably also abbreviations) as one of the several means he had at his disposal for absorbing extra text, but that he did it occasionally, and only as a last resource when the text that needed to be absorbed was not considerably long. To compensate for the numerous expansions he made, apprentice Y had to set his lines of type tighter than was his normal practice, but he nevertheless set a substantial number of quads before punctuation marks (thirty-nine; an average of ten quads per page), showing once again that this habit was one of his stronger compositorial trademarks. It is evident, too, that his spelling preferences agreed almost one hundred per cent with those of compositor K's. In other words, the changes introduced in the re-set pages and the lack of orthographic variants between the original and the re-set pages corroborate and reinforce several of the theories stated earlier.

Fascinating as the study of these typographical and textual characteristics of the finished book might be, the most crucial questions still remain to be answered. Did compositor K and apprentice Y respect the orthography of their copy? Does the remarkable spelling consistency we find in the first edition of Novelas ejemplares indicate that these Cuesta men retained most of the authorial spellings? Or, on the contrary, should it be taken as a crystal clear sign that they imposed their own preferred spellings relentlessly over and above those of the original manuscript?

The striking consistency of the compositors in their use of the form vuessa merced and its abbreviations (v.m., vs.ms.) suggests that they followed the authorial spellings conscientiously. But knowing, as we now know, that Cervantes spelled the past indicative and the future and past subjunctive of the auxiliary verb haber without initial "h" (vuo, vuiere, vuiera, vuiesse; see The Compositors, especially pp. 88-89), and that he used both forms of the word misma (misma and mesma), it comes as a surprise to find out that both compositor K and apprentice Y set all occurrences of these forms of the auxiliary verb with initial "h" (huuo, huuiere, huuiera, huuiesse), and that no occurrences of the form mesma could be found in Novelas ejemplares.

It follows, then, that compositor K and apprentice Y used only the form vuessa merced (vis-à-vis vuestra merced) solely because their preferred spelling agreed in this instance with that of Cervantes's, not because they accepted or were following the original orthography of the manuscript. When their preferred spellings did not agree with those of Cervantes's, these Cuesta men did not hesitate in the least to impose their own preferred spellings or

to regularize what had intentionally been spelled differently for purposes stylistic or for characterization (see above, estoria / historia, footnote 7).

The word Constantinopla, amongst dozens of other words I could bring forth, furnishes an excellent example of these workmen's tampering with the original orthography. In the first edition of Don Quixote, Part I, we find this word spelled either Costantinopla (pages Ff8v, Hh4v, Kk3, and Kk3v) or Constantinopla (pages Gg1, Gg2v, Gg3[2], Gg4, and Gg4v[2]). Compositors D and E set the form Costantinopla. Compositor F used the other form. But when compositor E set all these occurrences in the second Madrid edition of Don Quixote, Part I, he set them exactly as they appear in the first edition, an example of which he used as his copy, i.e., he set four occurrences of the form Costantinopla and seven occurrences of the form Constantinopla. Hence we can conclude 1) that compositor E was a reliable workman because, even though he might have preferred one spelling to the other, he nevertheless followed the orthography of his copy with some care, and thus, 2) that Cervantes's spelling (and pronunciation!) was Costantinopla. And yet, of the thirteen occurrences of this word set in the first edition of Novelas ejemplares (pages G1, G1v, G3, G4[2], G5v, G6, H2v[2], H3[3], and H6) I found no occurrences of the authorial form. This fact clearly shows that compositor K and apprentice Y had little respect for the original orthography of their copy.

Summing up: the setting and printing of the first edition of Novelas ejemplares was a relatively uneventful job which progressed steadily though at an accelerated pace with the pressmen running off the formes as soon as they left the hands of the compositors.[22] Of the five hundred and seventy pages with text (¶lv blank; Mm2v, colophon) only pages C3, C3v, C4, C4v, C5, C5v, C6, C6v, E4, E4v, E5, and E5v were set from cast-off copy. Four other pages (Ee4, Ee4v, Ee5, and Ee5v) had to be set and run off twice due to some accident that damaged one half of the inner sheet of gathering Ee of a considerable number of sheets. Therefore the copies of this edition should be separated into two family groups: those copies having the pages of the first printing (like copy G.10181, The British Library) belong to family group A, and those having the re-set pages (like copy C.59.b.20, The British Library, and the copy of The Hispanic Society of America) belong to family group B.

The setting of the entire work was assigned to one workman, compositor K, but the text in fact was set by two Cuesta men, compositor K and his apprentice. The orthography and setting habits of these two men were very similar, a circumstance that was to be expected. Apprentice Y was learning

the trade under the direction of compositor K and, naturally, had to accept and follow closely the grammatical rules laid down by, and the spelling preferences of, his instructor, and in the process acquire similar setting habits. But a dozen key words and the distinctive quadmarks of these Cuesta men served as the basis for separating and identifying them, and also for giving us an approximate idea of the pages, or sections of pages each workman set. More important, however, is the fact that these Cuesta men changed many of the authorial spellings relentlessly. After studying the spellings used in the first edition of Novelas ejemplares and comparing them with the spellings used in the first editions of Galatea, Don Quixote, Persiles y Sigismunda, Ocho comedias, and Viaje del Parnaso, we can conclude, albeit tentatively, that the text we have in the first edition of Novelas ejemplares is probably the most corrupt text we have of any of Cervantes's works. And thus, what initially might have been seen as merely an academic exercise acquires a new and more urgent character because we now realize that the attitude that compositor K and apprentice Y had towards the orthography of their copy reached at the very roots of Cervantes's style and lexicon.

As long as we did not know that the apparently chaotic orthography we have in the first editions of Cervantes's works was the result of the compositors imposing their individual spelling and punctuation preferences over and above those of Cervantes's, eclectic editing and regularization were in some degree justified. Now, however, we know not only that the first editions reflect the original orthography only vaguely, but also that it is possible to recover many of the authorial spellings and linguistic nuances of Cervantes's style. Editors of Cervantes's works have in front of them an unexplored and unexploited gold mine. My own research on the writing and on the setting and printing of the first editions of Don Quixote and Novelas ejemplares, and the conclusions reached therein, should not be accepted without first putting them through a systematic, rigorous, and thorough testing. Editors and critics should query and double check every one of my findings and theories to ascertain the reliability of my conclusions, to correct my mistakes, and to propose different or more refined solutions than I have. Only through this humanistic process of constant questioning, testing, revising, and improving can we hope to produce reliable editions of our texts. I personally think that a definitive edition is an impossible dream, a contradiction of terms, but editors should rise to the challenge and quixotically attempt to achieve the unattainable.