LETTERS OF
PLINY THE YOUNGER

T HE Pierpont Morgan Library, itself a work of art, contains masterpieces of painting and sculpture, rare books, and illuminated manuscripts. Scholars generally are perhaps not aware that it also possesses the oldest Latin manuscripts in America, including several that even the greatest European libraries would be proud to own. The collection is also admirably representative of the development of script throughout the Middle Ages. It comprises specimens of the uncial hand, the half-uncial, the Merovingian minuscule of the Luxeuil type, the script of the famous school of Tours, the St. Gall type, the Irish and Visigothic hands, and the Beneventan and Anglo-Saxon scripts.

Among the oldest manuscripts of the library, in fact the oldest, is a hitherto unnoticed fragment of great significance not only to palaeographers, but to all students of the classics. It consists of six leaves of an early sixth-century manuscript of the Letters of the younger Pliny. This new witness to the text, older by three centuries than the oldest codex heretofore used by any modern editor, has reappeared in this unexpected quarter, after centuries of wandering and hiding. The fragment was bought by the late J. Pierpont Morgan in Rome, in December 1910, from the art dealer Imbert; he had obtained it from De Marinis, of Florence, who had it from the heirs of the Marquis Taccone, of Naples. Nothing is known of the rest of the manuscript.

The present writers had the good fortune to visit the Pierpont Morgan Library in 1915. One of the first manuscripts put into their hands was this early sixth-century fragment of Pliny’s Letters, which forms the subject of the following pages. Having received permission to study the manuscript and publish results, they lost no time in acquainting classical scholars with this important find. In December of the same year, at the joint meeting of the American Archaeological and Philological Associations, held at Princeton University, two papers were read, one concerning the palaeographical, the other the textual, importance of the fragment. The two studies which follow, Part I by Doctor Lowe, Part II by Professor Rand, are an elaboration of the views presented at the meeting. Some months after the present volume was in the form of page-proof, Professor E. T. Merrill’s long-expected edition of Pliny’s Letters appeared (Teubner, Leipsic, 1922). We regret that we could not avail ourselves of it in time to introduce certain changes. The reader will still find Pliny cited by the pages of Keil, and in general he should regard the date of our production as 1921 rather than 1922.

iv The writers wish to express their gratitude for the privilege of visiting the Pierpont Morgan Library and making full use of its facilities. For permission to publish the manuscript they are indebted to the generous interest of Mr. J. Pierpont Morgan. They also desire to make cordial acknowledgment of the unfailing courtesy and helpfulness of the Librarian, Miss Belle da Costa Greene, and her assistant, Miss Ada Thurston. Lastly, the writers wish to thank the Carnegie Institution of Washington for accepting their joint study for publication and for their liberality in permitting them to give all the facsimiles necessary to illustrate the discussion.

E. K. RAND.
E. A. LOWE.

Contents
 size
 vellum
 binding T HE Morgan fragment of Pliny the Younger contains the end of Book II and the beginning of Book III of the Letters (II, xx. 13-III, v. 4). The fragment consists of six vellum leaves, or twelve pages, which apparently formed part of a gathering or quire of the original volume.

The leaves measure 11-3/8 by 7 inches (286 x 180 millimeters); the written space measures 7-1/4 by 4-3/8 inches (175 x 114 millimeters); outer margin, 1-7/8 inches (50 millimeters); inner, 3/4 inch (18 millimeters); upper margin, 1-3/4 inches (45 millimeters); lower, 2-1/4 inches (60 millimeters).

The vellum is well prepared and of medium thickness. The leaves are bound in a modern pliable vellum binding with three blank vellum fly-leaves in front and seven in back, all modern. On the inside of the front cover is the book-plate of John Pierpont Morgan, showing the Morgan arms with the device: Onward and Upward. Under the book-plate is the press-mark M.462.

Ruling There are twenty-seven horizontal lines to a page and two vertical bounding lines. The lines were ruled with a hard point on the flesh side, each opened sheet being ruled separately: 48^v and 53^r, 49^r and 52^v, 50^v and 51^r. The horizontal lines were guided by knife-slits made in the outside margins quite close to the text space; the two vertical lines were guided by two slits in the upper margin and two in the lower. The horizontal lines were drawn across the open sheets and extended occasionally beyond the slits, more often just beyond the perpendicular bounding lines. The written space was kept inside the vertical bounding lines except for the initial letter of each epistle; the first letter of the address and the first letter of the epistle proper projected into the left margin. Here and there the scribe transgressed beyond the bounding line. On the whole, however, he observed the limits and seemed to prefer to leave a blank before the bounding line rather than to crowd the syllable into the space or go beyond the vertical line.

Relation of the six leaves to the rest of the manuscript One might suppose that the six leaves once formed a complete gathering of the original book, especially as the first and last pages, folios 48^r and 53^v have a darker appearance, as though they had been the outside leaves of a gathering that had been affected by exposure. But this darker appearance is sufficiently accounted for by the fact that both pages are on the hair side of the parchment, and the 4 hair side is always darker than the flesh side. Quires of six leaves or trinions are not unknown. Examples of them may be found in our oldest manuscripts. But they are the exception.¹ The customary quire is a gathering of eight leaves, forming a quaternion proper. It would be natural, therefore, to suppose that our fragment did not constitute a complete gathering in itself but formed part of a quaternion. The supposition is confirmed by the following considerations:

In the first place, if our six leaves were once a part of a quaternion, the two leaves needed to complete them must have formed the outside sheet, since our fragment furnishes a continuous text without any lacuna whatever. Now, in the formation of quires, sheets were so arranged that hair side faced hair side, and flesh side flesh side. This arrangement is dictated by a sense of uniformity. As the hair side is usually much darker than the flesh side the juxtaposition of hair and flesh sides would offend the eye. So, in the case of our six leaves, folios 48^v and 53^r, presenting the flesh side, face folios 49^r and 52^v likewise on the flesh side; and folios 49^v and 52^r presenting the hair side, face folios 50^r and 51^v likewise on the hair side. The inside pages 50^v and 51^r which face each other, are both flesh side, and the outside pages 48^r and 53^v are both hair side, as may be seen from the accompanying diagram.

From this arrangement it is evident that if our fragment once formed part of a quaternion the missing sheet was so folded that its hair side faced the present outside sheet and its flesh side was on the outside of the whole gathering. Now, it was by far the more usual practice in our oldest uncial manuscripts to have the flesh side on the outside of the quire.² And as our fragment belongs to the oldest 5 class of uncial manuscripts, the manner of arranging the sheets of quires seems to favor the supposition that two outside leaves are missing. The hypothesis is, moreover, strengthened by another consideration. According to the foliation supplied by the fifteenth-century Arabic numerals, the leaf which must have followed our fragment bore the number 54, the leaf preceding it having the number 47. If we assume that our fragment was a complete gathering, we are obliged to explain why the next gathering began on a leaf bearing an even number (54), which is abnormal. We do not have to contend with this difficulty if we assume that folios 47 and 54 formed the outside sheet of our fragment, for six quires of eight leaves and one of six would give precisely 54 leaves. It seems, therefore, reasonable to assume that our fragment is not a complete unit, but formed part of a quaternion, the outside sheet of which is missing.

Original size of the manuscript In the fifteenth century, as the previous demonstration has made clear, our fragment was preceded by 47 leaves that are missing to-day. With this clue in our possession it can be demonstrated that the manuscript began with the first book of the Letters. We start with the fact that not all the 47 folios (or 94 pages) which preceded our six leaves were devoted to the text of the Letters. For, from the contents of our six leaves we know that each book must have been preceded by an index of addresses and first lines. The indices for Books I and II, if arranged in general like that of Book III, must have occupied four pages.³ We also learn from our fragment that space must be allowed for a colophon at the end of each book. One page for the colophons of Books I and II is a reasonable allowance. Accordingly it follows that out of the 94 pages preceding our fragment 5 were not devoted to text, or in other words that only 89 pages were thus devoted.

Now, if we compare pages in our manuscript with pages of a printed text we find that the average page in our manuscript corresponds to about 19 lines of the Teubner edition of 1912. If we multiply 89 by 19 we get 1691. This number of lines of the size of the Teubner edition should, if our calculation be correct, contain the text of the Letters preceding our fragment. The average page of the Teubner edition of 1912 of the part which interests us contains a little over 29 lines. If we divide 1691 by 29 we get 58.3. Just 58 pages of Teubner text are occupied by the 47 leaves which preceded our fragment. So close a conformity is sufficient to prove our point. We have possibly allowed too much space for indices and colophons, especially if the former covered less ground for 6 Books I and II than for Book III. Further, owing to the abbreviation of que and bus, and particularly of official titles, we can not expect a closer agreement.

It is not worth while to attempt a more elaborate calculation. With the edges matching so nearly, it is obvious that the original manuscript as known and used in the fifteenth century could not have contained some other work, however brief, before Book I of Pliny’s Letters. If the manuscript contained the entire ten books it consisted of about 260 leaves. This sum is obtained by counting the number of lines in the Teubner edition of 1912, dividing this sum by 19, and adding thereto pages for colophons and indices. It would be too bold to suppose that this calculation necessarily gives us the original size of the manuscript, since the manuscript may have had less than ten books, or it may, on the other hand, have had other works. But if it contained only the ten books of the Letters, then 260 folios is an approximately correct estimate of its size.

It is hard to believe that only six leaves of the original manuscript have escaped destruction. The fact that the outside sheet (foll. 48^r and 53^v) is not much worn nor badly soiled suggests that the gathering of six leaves must have been torn from the manuscript not so very long ago and that the remaining portions may some day be found.

Disposition The pages in our manuscript are written in long lines,⁴ in scriptura continua, with hardly any punctuation.

Each page begins with a large letter, even though that letter occur in the body of a word (cf. foll. 48^r, 51^v, 52^r).⁵

Each epistle begins with a large letter. The line containing the address which precedes each epistle also begins with a large letter. In both cases the large letter projects into the left margin.

The running title at the top of each page is in small rustic capitals.⁶ On the verso of each folio stands the word EPISTVLARVM; on the recto of the following folio stands the number of the book, e.g., LIB. II, LIB. III.

To judge by our fragment, each book was preceded by an index of addresses 7 and initial lines written in alternating lines of black and red uncials. Alternating lines of black and red rustic capitals of a large size were used in the colophon.⁷

Ornamentation As in all our oldest Latin manuscripts, the ornamentation is of the simplest kind. Such as it is, it is mostly found at the end and beginning of books. In our case, the colophon is enclosed between two scrolls of vine-tendrils terminating in an ivy-leaf at both ends. The lettering in the colophon and in the running title is set off by means of ticking above and below the line.

Red is used for decorative purposes in the middle line of the colophon, in the scroll of vine-tendrils, in the ticking, and in the border at the end of the Index on fol. 49. Red was also used, to judge by our fragment, in the first three lines of a new book,⁸ in the addresses in the Index, and in the addresses preceding each letter.

Corrections The original scribe made a number of corrections. The omitted line of the Index on fol. 49 was added between the lines, probably by the scribe himself, using a finer pen; likewise the omitted line on fol. 52^v, lines 7-8. A number of slight corrections come either from the scribe or from a contemporary reader; the others are by a somewhat later hand, which is probably not more recent than the seventh century.⁹ The method of correcting varies. As a rule, the correct letter is added above the line over the wrong letter; occasionally it is written over an erasure. An omitted letter is also added above the line over the space where it should be inserted. Deletion of single letters is indicated by a dot placed over the letter and a horizontal or an oblique line drawn through it. This double use of expunction and cancellation is not uncommon in our oldest manuscripts. For details on the subject of corrections, see the notes on pp. 23-34.

There is a ninth-century addition on fol. 53 and one of the fifteenth century on fol. 51. On fol. 49, in the upper margin, a fifteenth-century hand using a stilus or hard point scribbled a few words, now difficult to decipher.¹⁰ Presumably the same hand drew a bearded head with a halo. Another relatively recent hand, using lead, wrote in the left margin of fol. 53^v the monogram QR¹¹ and the roman numerals i, ii, iii under one another. These numerals, as Professor 8 Rand correctly saw, refer to the works of Pliny the Elder enumerated in the text. Further activity by this hand, the date of which it is impossible to determine, may be seen, for example, on fol. 49^v, ll. 8, 10, 15; fol. 52, ll. 4, 10, 13, 21, 22; fol. 53, ll. 12, 15, 16, 17, 20, 27; fol. 53^v, ll. 5, 10, 15.

Syllabification Syllables are divided after a vowel or diphthong except where such a division involves beginning the next syllable with a group of consonants.¹² In that case the consonants are distributed between the two syllables, one consonant going with one syllable and the other with the following, except when the group contains more than two successive consonants, in which case the first consonant goes with the first syllable, the rest with the following syllable. That the scribe is controlled by this mechanical rule and not by considerations of pronunciation is obvious from the division san|ctissimum and other examples found below. The method followed by him is made amply clear by the examples which occur in our twelve pages:¹³

Orthography The spelling found in our six leaves is remarkably correct. It compares favorably with the best spelling encountered in our oldest Latin manuscripts of the fourth and fifth centuries. The diphthong ae is regularly distinguished from e. The interchange of b and u, d and t, o and u, so common in later manuscripts, is rare here: the confusion between b and u occurs once (comprouasse, fo. 52^v, l. 1); the omission of h occurs once (pulcritudo, fo. 51^v, l. 26); the use of k for c occurs twice (karet, fo. 51^r, l. 14, and karitas, fo. 52^r, l. 5). The scribe uses the correct forms in adolescet (fo. 51^v, l. 14) and adulescenti (fo. 51^v, l. 24); he writes auonculi (fo. 53^v, l. 15), exsistat (fo. 51^v, l. 9), and exsecutos (fo. 53^r, l. 8). In the case of composite words he has the assimilated form in some, and in others the unassimilated form, as the following examples go to show:

Abbreviations Very few abbreviated words occur in our twelve pages. Those that are found are subject to strict rules. What is true of the twelve pages was doubtless true of the entire manuscript, inasmuch as the sparing use of abbreviations in conformity with certain definite rules is a characteristic of all our oldest manuscripts.¹⁴ The abbreviations found in our fragment may conveniently be grouped as follows:

1. Suspensions which might occur in any ancient manuscript or inscription, e.g.:

2. Technical or recurrent terms which occur in the colophons at the end of each book and at the end of letters, as:

4. Omitted M at the end of a line, omitted N at the end of a line, the omission being indicated by means of a horizontal stroke, thickened at either end, which is placed over the space immediately following the final vowel.¹⁹ This omission may occur in the middle of a word but only at the end of a line.

Authenticity of the six leaves The sudden appearance in America of a portion of a very ancient classical manuscript unknown to modern editors may easily arouse suspicion in the minds of some scholars. Our experience with the “Anonymus Cortesianus” has taught us to be wary,²⁰ and it is natural to demand proof establishing the genuineness of the new fragment.²¹ As to the six leaves of the Morgan Pliny, it may be said unhesitatingly that no one with experience of ancient Latin manuscripts could entertain any doubt as to their genuineness. The look and feel of the parchment, the ink, the script, the titles, colophons, ornamentation, corrections, and later additions, all bear the indisputable marks of genuine antiquity.

But it may be objected that a clever forger possessing a knowledge of palaeography would be able to reproduce all these features of ancient manuscripts. This objection can hardly be sustained. It is difficult to believe that any modern could reproduce faithfully all the characteristics of sixth-century uncials and fifteenth-century notarial writing without unconsciously falling into some error and betraying his modernity. Besides, there is one consideration which to my mind establishes the genuineness of our fragment beyond a peradventure. We have seen above that the leaves of our manuscript are so arranged that hair side faces hair side and flesh side faces flesh side. The visible effect of this arrangement is that two pages of clear writing alternate with two pages of faded writing, the faded appearance being caused by the ink scaling off from the less porous surface of the flesh side of the vellum.²² As a matter of fact, the flesh side of the vellum showed 12 faded writing long before modern time. To judge by the retouched characters on fol. 53^r it would seem that the original writing had become illegible by the eighth or ninth century.²³ Still, a considerable period of time would, so far as we know, be necessary for this process. It is highly improbable that a forger could devise this method of giving his forgery the appearance of antiquity, and even if he attempted it, it is safe to say that the present effect would not be produced in the time that elapsed before the book was sold to Mr. Morgan.

But let us assume, for the sake of argument, that the Morgan fragment is a modern forgery. We are then constrained to credit the forger not only with a knowledge of palaeography which is simply faultless, but, as will be shown in the second part, with a minute acquaintance with the criticism and the history of the text. And this forger did not try to attain fame or academic standing by his nefarious doings, as was the case with the Roman author of the forged “Anonymus Cortesianus,” for nothing was heard of this Morgan fragment till it had reached the library of the American collector. If his motive was monetary gain he chose a long and arduous path to attain it. It is hardly conceivable that he should take the trouble to make all the errors and omissions found in our twelve pages and all the additions and corrections representing different ages, different styles, when less than half the number would have served to give the forged document an air of verisimilitude. The assumption that the Morgan fragment is a forgery thus becomes highly unreasonable. When you add to this the fact that there is nothing in the twelve pages that in any way arouses suspicion, the conclusion is inevitable that the Morgan fragment is a genuine relic of antiquity.

Archetype As to the original from which our manuscript was copied, very little can be said. The six leaves before us furnish scanty material on which to build any theory. The errors which occur are not sufficient to warrant any conclusion as to the script of the archetype. One item of information, however, we do get: an omission on fol. 52^v goes to show that the manuscript from which our scribe copied was written in lines of 25 letters or thereabout.²⁴ The scribe first wrote excucuris|sem commeatu. Discovering his error of omission, he erased sem at the beginning of line 8 and added it at the end of line 7 (intruding upon margin-space in order to do so), and then supplied, in somewhat smaller letters, the omitted words accepto ut praefectus aerari. As there are no homoioteleuta to 13 account for the omission, it is almost certain that it was caused by the inadvertent skipping of a line.²⁵ The omitted letters number 25.

A glance at the abbreviations used in the index of addresses on foll. 48^v-49^r teaches that the original from which our manuscript was copied must have had its names abbreviated in exactly the same form. There is no other way of explaining why the scribe first wrote ad iulium seruianum (fol. 49, l. 12), and then erased the final um and put a point after seruian.

Our manuscript was written in Italy at the end of the fifth or more probably at the beginning of the sixth century.

The manuscripts with which we can compare it come, with scarcely an exception, from Italy; for it is only of more recent uncial manuscripts (those of the seventh and eighth centuries) that we can say with certainty that they originate in other than Italian centres. The only exception which occurs to one is the Codex Bobiensis (k) of the Gospels of the fifth century, which may actually have been written in Africa, though this is far from certain. As for our fragment, the details of its script, as well as the ornamentation, disposition of the page, the ink, the parchment, all find their parallels in authenticated Italian products; and this similarity in details is borne out by the general impression of the whole.

The manuscript may be dated at about the year A.D. 500, for the reason that the script is not quite so old as that of our oldest fifth-century uncial manuscripts, and yet decidedly older than that of the Codex Fuldensis of the Gospels (F) written in or before A.D. 546.

On the dating of uncial manuscripts In dating uncial manuscripts we must proceed warily, since the data on which our judgments are based are meagre in the extreme and rather difficult to formulate.

The history of uncial writing still remains to be written. The chief value of excellent works like Chatelain’s Uncialis Scriptura or Zangemeister and Wattenbach’s Exempla Codicum Latinorum Litteris Maiusculis Scriptorum lies in the mass of material they offer to the student. This could not well be otherwise, since clear-cut, objective criteria for dating uncial manuscripts have not yet been formulated; and that is due to the fact that of our four hundred or more uncial manuscripts, ranging from the fourth to the eighth century, very few, indeed, can be dated with 14 precision, and of these virtually none is in the oldest class. Yet a few guide-posts there are. By means of those it ought to be possible not only to throw light on the development of this script, but also to determine the features peculiar to the different periods of its history. This task, of course, can not be attempted here; it may, however, not be out of place to call attention to certain salient facts.

The student of manuscripts knows that a law of evolution is observable in writing as in other aspects of human endeavor. The process of evolution is from the less to the more complex, from the less to the more differentiated, from the simple to the more ornate form. Guided by these general considerations, he would find that his uncial manuscripts naturally fall into two groups. One group is manifestly the older: in orthography, punctuation, and abbreviation it bears close resemblance to inscriptions of the classical or Roman period. The other group is as manifestly composed of the more recent manuscripts: this may be inferred from the corrupt or barbarous spelling, from the use of abbreviations unfamiliar in the classical period but very common in the Middle Ages, or from the presence of punctuation, which the oldest manuscripts invariably lack. The manuscripts of the first group show letters that are simple and unadorned and words unseparated from each other. Those of the second group show a type of ornate writing, the letters having serifs or hair-lines and flourishes, and the words being well separated. There can be no reasonable doubt that this rough classification is correct as far as it goes; but it must remain rough and permit large play for subjective judgement.

A scientific classification, however, can rest only on objective criteria—criteria which, once recognized, are acceptable to all. Such criteria are made possible by the presence of dated manuscripts. Now, if by a dated manuscript we mean a manuscript of which we know, through a subscription or some other entry, that it was written in a certain year, there is not a single dated manuscript in uncial writing which is older than the seventh century—the oldest manuscript with a precise date known to me being the manuscript of St. Augustine written in the Abbey of Luxeuil in A.D. 669.²⁶ But there are a few manuscripts of which we can say with certainty that they were written either before or after some given date. And these manuscripts which furnish us with a terminus ante quem or post quem, as the case may be, are extremely important to us as being the only relatively safe landmarks for following development in a field that is both remote and shadowy.

The Codex Fuldensis of the Gospels, mentioned above, is our first landmark of importance.²⁷ It was read by Bishop Victor of Capua in the years A.D. 546 and 547, as is testified by two entries, probably autograph. From this it follows that 15 the manuscript was written before A.D. 546. We may surmise—and I think correctly—that it was shortly before 546, if not in that very year. In any case the Codex Fuldensis furnishes a precise terminus ante quem.

The other landmark of importance is furnished by a Berlin fragment containing a computation for finding the correct date for Easter Sunday.²⁸ Internal evidence makes it clear that this Computus Paschalis first saw light shortly after A.D. 447. The presumption is that the Berlin leaves represent a very early copy, if not the original, of this composition. In no case can these leaves be regarded as a much later copy of the original, as the following purely palaeographical considerations, that is, considerations of style and form of letters, will go to show.

Let us assume, as we do in geometry, for the sake of argument, that the Fulda manuscript and the Berlin fragment were both written about the year 500—a date representing, roughly speaking, the middle point in the period of about one hundred years which separates the extreme limits of the dates possible for either of these two manuscripts, as the following diagram illustrates:

If our hypothesis be correct, then the script of these two manuscripts, as well as other palaeographical features, would offer striking similarities if not close resemblance. As a matter of fact, a careful comparison of the two manuscripts discloses differences so marked as to render our assumption absurd. The Berlin fragment is obviously much older than the Fulda manuscript. It would be rash to specify the exact interval of time that separates these two manuscripts, yet if we remember the slow development of types of writing the conclusion seems justified that at least several generations of evolution lie between the two manuscripts. If this be correct, we are forced to push the date of each as far back as the ascertained limit will permit, namely, the Fulda manuscript to the year 546 and the Berlin fragment to the year 447. Thus, apparently, considerations of form and style (purely palaeographical considerations) confirm the dates derived from examination of the internal evidence, and the Berlin and Fulda manuscripts may, in effect, be considered two dated manuscripts, two definite guide-posts.

If the preceding conclusion accords with fact, then we may accept the traditional date (circa A.D. 371) of the Codex Vercellensis of the Gospels. The famous Vatican palimpsest of Cicero’s De Re Publica seems more properly placed in the fourth than in the fifth century; and the older portion of the Bodleian manuscript of Jerome’s translation of the Chronicle of Eusebius, dated after the year A.D. 442, becomes another guide-post in the history of uncial writing, since a comparison with the Berlin fragment of about A.D. 447 convinces 16 one that the Bodleian manuscript can not have been written much after the date of its archetype, which is A.D. 442.

Dated uncial manuscripts Asked to enumerate the landmarks which may serve as helpful guides in uncial writing prior to the year 800, we should hardly go far wrong if we tabulate them in the following order:²⁹

ca. a. 371 1. Codex Vercellensis of the Gospels (a).

post a. 442 2. Bodleian Manuscript (Auct. T. 2. 26) of Jerome’s translation of the Chronicle of Eusebius (older portion).

ca. a. 447 3. Berlin Computus Paschalis (MS. lat. 4º. 298).

ante a. 546 4. Codex Fuldensis of the Gospels (F), Fulda MS. Bonifat. 1, read by Bishop Victor of Capua.

a. 438-ca. 550 5. Codex Theodosianus (Turin, MS. A. II. 2).

Manuscripts containing the Theodosian Code can not be earlier than A.D. 438, when this body of law was promulgated, nor much later than the middle of sixth century, when the Justinian Code supplanted the Theodosian and made it useless to copy it.

a. 600-666 6. The Toulouse Manuscript (No. 364) and Paris MS. lat. 8901, containing Canons, written at Albi.

a. 669 7. The Morgan Manuscript of St. Augustine’s Homilies, written in the Abbey of Luxeuil. Later at Beauvais and Chateau de Troussures.

a. 699 8. The Berne Manuscript (No. 219B) of Jerome’s translation of the Chronicle of Eusebius, written in France, possibly at Fleury.

a. 695-711 9. Brussels Fragment of a Psalter and Varia Patristica (MS. 1221 = 9850-52) written for St. Medardus in Soissons in the time of Childebert III.

ante a. 716 10. Codex Amiatinus of the Bible (Florence Laur. Am. 1) written in England.

a. 719 11. The Treves Prosper (MS. 36, olim S. Matthaei).

ca. a. 750 12. The Milan Manuscript (Ambros. B. 159 sup.) of Gregory’s Moralia, written at Bobbio in the abbacy of Anastasius.

ante a. 752 13. The Bodleian Acts of the Apostles (MS. Selden supra 30) written in the Isle of Thanet.

a. 754 14. The Autun Manuscript (No. 3) of the Gospels, written at Vosevium.

a. 739-760 15. Codex Beneventanus of the Gospels (London Brit. Mus. Add. MS. 5463) written at Benevento.

post a. 787 16. The Lucca Manuscript (No. 490) of the Liber Pontificalis.

Guided by the above manuscripts, we may proceed to determine the place which the Morgan Pliny occupies in the series of uncial manuscripts. The student of manuscripts recognizes at a glance that the Morgan fragment is, as has been said, distinctly older than the Codex Fuldensis of about the year 546. But how much older? Is it to be compared in antiquity with such venerable monuments as the palimpsest of Cicero’s De Re Publica, with products like the Berlin Computus Paschalis or the Bodleian Chronicle of Eusebius? If we examine carefully the characteristics of our oldest group of fourth- and fifth-century manuscripts and compare them with those of the Morgan manuscript we shall see that the latter, though sharing some of the features found in manuscripts of the oldest group, lacks others and in turn shows features peculiar to manuscripts of a later group.

Oldest group of uncial manuscripts Our oldest group would naturally be composed of those uncial manuscripts which bear the closest resemblance to the above-mentioned manuscripts of the fourth and fifth centuries, and I should include in that group such manuscripts as these:

1. Rome, Vatic. lat. 5757.—Cicero, De Re Publica, palimpsest.

2. Rome, Vatic. lat. 5750 + Milan, Ambros. E. 147 sup.—Scholia Bobiensia in Ciceronem, palimpsest.

3. Vienna, 15.—Livy, fifth decade (five books).

4. Paris, lat. 5730.—Livy, third decade.

5. Verona, XL (38).—Livy, first decade, 6 palimpsest leaves.

6. Rome, Vatic. lat. 10696.—Livy, fourth decade, Lateran fragments.

7. Bamberg, Class. 35a.—Livy, fourth decade, fragments.

8. Vienna, lat. 1a.—Pliny, Historia Naturalis, fragments.

9. St. Paul in Carinthia, XXV a 3.—Pliny, Historia Naturalis, palimpsest.

10. Turin, A. II. 2.—Theodosian Codex, fragments, palimpsest.

1. Vercelli, Cathedral Library.—Gospels (a) ascribed to Bishop Eusebius (†371).

2. Paris, lat. 17225.—Corbie Gospels (ff²).

3. Constance-Weingarten Biblical fragments.—Prophets, fragments scattered in the libraries of Stuttgart, Darmstadt, Fulda, and St. Paul in Carinthia.

4. Berlin, lat. 4º. 298.—Computus Paschalis of ca. a. 447.

5. Turin, G. VII. 15.—Bobbio Gospels (k).

6. Turin, F. IV. 27 + Milan, D. 519. inf. + Rome, Vatic. lat. 10959.—Cyprian, Epistolae, fragments.

7. Turin, G. V. 37.—Cyprian, de opere et eleemosynis.

8. Oxford, Bodleian Auct. T. 2. 26.—Eusebius-Hieronymus, Chronicle, post a. 442.

9. Petrograd Q. v. I. 3 (Corbie).—Varia of St. Augustine.

10. St. Gall, 1394.—Gospels (n).

Characteristics of the oldest uncial manuscripts The main characteristics of the manuscripts included in the above list, which is by no means complete, may briefly be described thus:

1. General effect of compactness. This is the result of scriptura continua, which knows no separation of words and no punctuation. See the facsimiles cited above.

2. Precision in the mode of shading. The alternation of stressed and unstressed strokes is very regular. The two arcs of are shaded not in the middle, as in Greek uncials, but in the lower left and upper right parts of the letter, so that the space enclosed by the two arcs resembles an ellipse leaning to the left at an angle of about 45°, thus . What is true of the is true of other curved strokes. The strokes are often very short, mere touches of pen to parchment, like brush work. Often they are unconnected, thus giving a mere suggestion of the form. The attack or fore-stroke as well as the finishing stroke is a very fine, oblique hair-line.³⁰

3. Absence of long ascending or descending strokes. The letters lie virtually between two lines (instead of between four as in later uncials), the upper and lower shafts of letters like projecting but slightly beyond the head and base lines.

4. The broadness of the letters

5. The relative narrowness of the letters

6. The manner of forming

B with the lower bow considerably larger than the upper, which often has the form of a mere comma.

E with the tongue or horizontal stroke placed not in the middle, as in later uncial manuscripts, but high above it, and extending beyond the upper curve. The loop is often left open.

L with very small base.

M with the initial stroke tending to be a straight line instead of the well-rounded bow of later uncials.

N with the oblique connecting stroke shaded.

P with the loop very small and often open.

S with a rather longish form and shallow curves, as compared with the broad form and ample curves of later uncials.

T with a very small, sinuous horizontal top stroke (except at the beginning of a line when it often has an exaggerated extension to the left).

7. Extreme fineness of parchment, at least in parts of the manuscript.

20 8. Perforation of parchment along furrows made by the pen.

9. Quires signed by means of roman numerals often preceded by the letter Q· (= Quaternio) in the lower right corner of the last page of each gathering.

10. Running titles, in abbreviated form, usually in smaller uncials than the text.

11. Colophons, in which red and black ink alternate, usually in large-sized uncials.

12. Use of a capital, i.e., a larger-sized letter at the beginning of each page or of each column in the page, even if the beginning falls in the middle of a word.

13. Lack of all but the simplest ornamentation, e.g., scroll or ivy-leaf.

14. The restricted use of abbreviations. Besides B· and Q· and such suspensions as occur in classical inscriptions only the contracted forms of the Nomina Sacra are found.

15. Omission of M and N allowed only at the end of a line, the omission being marked by means of a simple horizontal line (somewhat hooked at each end) placed above the line after the final vowel and not directly over it as in later uncial manuscripts.

16. Absence of nearly all punctuation.

17. The use of in the text where an omission has occurred, and after the supplied omission in the lower margin, or the same symbols reversed if the supplement is entered in the upper margin.

If we now turn to the Morgan Pliny we observe that it lacks a number of the characteristics enumerated above as belonging to the oldest type of uncial manuscripts. The parchment is not of the very thin sort. There has been no corrosion along the furrows made by the pen. The running title and colophons are in rustic capitals, not in uncials. The manner of forming such letters as differs from that employed in the oldest group.

B with the lower bow not so markedly larger than the upper.

E with the horizontal stroke placed nearer the middle.

M with the left bow tending to become a distinct curve.

R S T have gained in breadth and proportionately lost in height.

Date of the Morgan manuscript Inasmuch as these palaeographical differences mark a tendency which reaches fuller development in later uncial manuscripts, it is clear that their presence in our manuscript is a sign of its more recent character as compared with manuscripts of the oldest type. Just as our manuscript is clearly older than the Codex Fuldensis of about the year 546, so it is clearly more recent than the Berlin Computus Paschalis of about the year 447. Its proper place is at the end of the oldest series of uncial manuscripts, which begins with the Cicero palimpsest. Its closest neighbors are, I believe, the Pliny palimpsest of St. Paul in Carinthia and the Codex Theodosianus of Turin. If we conclude by saying that the Morgan manuscript was written about the year 500 we shall probably not be far from the truth.

Later history of the Morgan manuscript The vicissitudes of a manuscript often throw light upon the history of the text contained in the manuscript. And the palaeographer knows that any scratch or scribbling, any probatio pennae or casual entry, may become important in tracing the wanderings of a manuscript.

In the six leaves that have been saved of our Morgan manuscript we have two entries. One is of a neutral character and does not take us further, but the other is very clear and tells an unequivocal story.

The unimportant entry occurs in the lower margin of folio 53^r. The words “uir erat in terra,” which are apparently the beginning of the book of Job, are written in Carolingian characters of the ninth century. As these characters were used during the ninth century in northern Italy as well as in France, it is impossible to say where this entry was made. If in France, then the manuscript of Pliny must have left its Italian home before the ninth century.³¹

That it had crossed the Alps by the beginning of the fifteenth century we know from the second entry. Nay, we learn more precise details. We learn that our manuscript had found a home in France, in the town of Meaux or its vicinity. The entry is found in the upper margin of fol. 51^r and doubtless represents a probatio pennae on the part of a notary. It runs thus:

“A tous ceulz qui ces presentes lettres verront et orront
Jehan de Sannemeres garde du scel de la provoste de
Meaulx & Francois Beloy clerc Jure de par le Roy
nostre sire a ce faire Salut sachient tuit que par.”

The above note is made in the regular French notarial hand of the fourteenth and fifteenth centuries.³² The formula of greeting with which the document opens is in the precise form in which it occurs in numberless charters of the period. All efforts to identify Jehan de Sannemeres, keeper of the seal of the provosté of Meaux, and François Beloy, sworn clerk in behalf of the King, have so far proved fruitless.³³

Conclusion Our manuscript, then, was written in Italy about the year 500. It is quite possible that it had crossed the Alps by the ninth century or even before. It is certain that by the fifteenth century it had found asylum in France. When and under what circumstances it got back to Italy will be shown by Professor Rand in the pages that follow.

So it is France that has saved this, the oldest extant witness of Pliny’s Letters, 22 for modern times. To mediaeval France we are, in fact, indebted for the preservation of more than one ancient classical manuscript. The oldest manuscript of the third decade of Livy was at Corbie in Charlemagne’s time, when it was loaned to Tours and a copy of it made there. Both copy and original have come down to us. Sallust’s Histories were saved (though not in complete form) for our generation by the Abbey of Fleury. The famous Schedae Vergilianae, in square capitals, as well as the Codex Romanus of Virgil, in rustic capitals, belonged to the monastery of St. Denis. Lyons preserved the Codex Theodosianus. It was again some French centre that rescued Pomponius Mela from destruction. The oldest fragments of Ovid’s Pontica, the oldest fragments of the first decade of Livy, the oldest manuscript of Pliny’s Natural History—all palimpsests—were in some French centre in the Middle Ages, as may be seen from the indisputably eighth-century French writing which covers the ancient texts. The student of Latin literature knows that the manuscript tradition of Lucretius, Suetonius, Cæsar, Catullus, Tibullus, and Propertius—to mention only the greatest names—shows that we are indebted primarily to Gallia Christiana for the preservation of these authors.

The Codex Parisinus A LDUS MANUTIUS, in the preface to his edition of Pliny’s Letters, printed at Venice in 1508, expresses his gratitude to Aloisio Mocenigo, Venetian ambassador in Paris, for bringing to Italy an exceptionally fine manuscript of the Letters; the book had been found not long before at or near Paris by the architect Fra Giocondo of Verona. The editio princeps, 1471, was based on a family of manuscripts that omitted Book VIII, called Book IX Book VIII, and did not contain Book X, the correspondence between Pliny and Trajan. Subsequent editions had only in part made good these deficiencies. More than a half of Book X, containing the letters numbered 41-121 in editions of our day, was published by Avantius in 1502 from a copy of the Paris manuscript made by Petrus Leander.² Aldus himself, two years before printing his edition, had received from Fra Giocondo a copy of the entire manuscript, with six other volumes, some of them printed editions which Giocondo had collated with manuscripts. Aldus, addressing Mocenigo, thus describes his acquisition:

“Deinde Iucundo Veronensi Viro singulari ingenio, ac bonarum literarum studiosissimo, quod et easdem Secundi epistolas ab eo ipso exemplari a se descriptas in Gallia diligenter ut facit omnia, et sex alia uolumina epistolarum partim manu scripta, partim impressa quidem, sed cum antiquis collata exemplaribus, ad me ipse sua sponte, quae ipsius est ergo studiosos omneis beneuolentia, adportauerit, idque biennio ante, quam tu ipsum mihi exemplar publicandum tradidisses.”

So now the ancient manuscript itself had come. Aldus emphasizes its value in supplying the defects of previous editions. The Letters will now include, he declares:

“multae non ante impressae. Tum Graeca correcta, et suis locis restituta, atque retectis adulterinis, uera reposita. Item fragmentatae epistolae, integrae factae. In medio etiam epistolae libri octaui de Clitumno fonte non solum uertici calx additus, et calci uertex, sed decem quoque epistolae interpositae, ac ex Nono libro Octauus factus, et ex Octauo Nonus, Idque beneficio exemplaris correctissimi, & mirae, ac uenerandae Vetustatis.”

The presence of such a manuscript, “most correct, and of a marvellous and venerable antiquity,” stimulates the imagination: Aldus thinks that now even the lost Decades of Livy may appear again:

“Solebam superioribus Annis Aloisi Vir Clariss. cum aut T. Liuii Decades, quae non extare creduntur, aut Sallustii, aut Trogi historiae, aut quemuis alium ex antiquis autoribus inuentum esse audiebam, nugas dicere, ac fabulas. Sed ex quo tu ex Gallia has Plinii epistolas in Italia reportasti, in membrana scriptas, atque adeo diuersis a nostris characteribus, ut nisi quis diu assuerit, non queat legere, coepi sperare mirum in modum, fore aetate nostra, ut plurimi ex bonis autoribus, quos non extare credimus, inueniantur.”

There was something unusual in the character of the script that made it hard to read; its ancient appearance even suggested to Aldus a date as early as that of Pliny himself.

“Est enim uolumen ipsum non solum correctissimum, sed etiam ita antiquum, ut putem scriptum Plinii temporibus.”

This is enthusiastic language. In the days of Italian humanism, a scholar might call almost any book a codex pervetustus if it supplied new readings for his edition and its script seemed unusual. As Professor Merrill remarks:³

“The extreme age that Aldus was disposed to attribute to the manuscript will, of course, occasion no wonder in the minds of those who are familiar with the vague notions on such matters that prevailed among scholars before the study of palaeography had been developed into somewhat of a science. The manuscript may have been written in one of the so-called ‘national’ hands, Lombardic, Visigothic, or Merovingian. But if it were in a ‘Gothic’ hand of the twelfth or thirteenth centuries, it might have appeared sufficiently grotesque and illegible to a reader accustomed for the most part to the exceedingly clear Italian book hands of the fifteenth century.”

In a later article Professor Merrill well adds that even the uncial script would have seemed difficult and alien to one accustomed to the current fifteenth-century style.⁴ A contemporary and rival editor, Catanaeus, disputed Aldus’s claims. In his second edition of the Letters (1518), he professed to have used a very ancient book that came down from Germany and declared that the Paris manuscript had no right to the antiquity which Aldus had imputed to it. But Catanaeus has been proved a liar.⁵ He had no ancient manuscript from Germany, and abused Aldus mainly to conceal his cribbings from that scholar’s edition; we may discount his opinion of the age of the Parisinus. Until Aldus, an eminent scholar and honest publisher,⁶ is proved guilty, we should assume him innocent of mendacity or naïve ignorance. He speaks in earnest; his words ring true. We must be prepared for the possibility that his ancient manuscript was really ancient.

39 Since Aldus’s time the Parisinus has disappeared. To quote Merrill again:⁷

“This wonderful manuscript, like so many others, appears to have vanished from earth. Early editors saw no especial reason for preserving what was to them but copy for their own better printed texts. Possibly some leaves of it may be lying hid in old bindings; possibly they went to cover preserve-jars, or tennis-racquets; possibly into some final dust-heap. At any rate the manuscript is gone; the copy by Iucundus is gone; the copy of the correspondence with Trajan that Avantius owed to Petrus Leander is gone; if others had any other copies of Book X, in whole or in part, they are gone too.”

The Bodleian volume In 1708 Thomas Hearne, the antiquary, bought at auction a peculiar volume of Pliny’s Letters. It consisted of Beroaldus’s edition of the nine books (1498), the portions of Book X published by Avantius in 1502, and, on inserted leaves, the missing letters of Books VIII and X.⁸ The printed portions, moreover, were provided with over five hundred variant readings and lemmata in a different hand from that which appeared on the inserted leaves; the hand that added the variants also wrote in the margin the sixteenth letter of Book IX, which is not in the edition of Beroaldus. Hearne recognized the importance of this supplementary matter, for he copied the variants into his own edition of the Letters (1703), intending, apparently, to use them in a larger edition which he is said to have published in 1709; he also lent the book to Jean Masson, who refers to it in his Plinii Vita. Upon Hearne’s death, this valuable volume was acquired by the Bodleian Library in Oxford, but lay unnoticed until Mr. E. G. Hardy, in 1888,⁹ examined it and, after a comparison of the readings, pronounced it the very copy from which Aldus had printed his edition in 1508. External proof of this highly exciting surmise seemed to appear in a manuscript note on the last page of the edition of Avantius, written in the hand that had inserted the variants and supplements throughout the volume:¹⁰

“hae plinii iunioris epistolae ex uetustissimo exemplari parisiensi et restitutae et emendatae sunt opera et industria ioannis iucundi prestantissimi architecti hominis imprimis antiquarii.”

What more natural to conclude than that here is the very copy that Aldus prepared from the ancient manuscript and the collations and transcripts sent him by Fra Giocondo? One fact blocks this attractive conjecture: though there are many agreements between the readings of the emended Bodleian book and those of Aldus, there are also many disagreements. Mr. Hardy removed the obstacle by assuming that Aldus made changes in the proof; but the changes are numerous; they are not too numerous for a scholar who can mark up his galleys free of cost, but they are decidedly too numerous if the scholar is also his own printer.

40 Merrill, in a brilliant and searching article,¹¹ entirely demolishes Hardy’s argument. Unlike most destructive critics, he replaces the exploded theory by still more interesting fact. For the rediscovery of the Bodleian book and a proper appreciation of its value, students of Pliny’s text must always be grateful to Hardy; we now know, however, that the volume was never owned by Aldus. The scholar who put its parts together and added the variants with his own hand was the famous Hellenist Guillaume Budé (Budaeus). The parts on the supplementary leaves were done by some copyist who imitated the general effect of the type used in the book itself; Budaeus added his notes on these inserted leaves in the same way as elsewhere. It had been shown before by Keil¹² that Budaeus must have used the readings of the Parisinus; indeed, it is from his own statement in Annotationes in Pandectas that we learn of the discovery of the ancient manuscript by Giocondo:¹³

“Verum haec epistola et aliae non paucae in codicibus impressis non leguntur: nos integrum ferme Plinium habemus: primum apud parrhisios repertum opera Iucundi sacerdotis: hominis antiquarii Architectique famigerati.”

The wording here is much like that in the note at the end of the Bodleian book. After establishing his case convincingly from the readings followed by Budaeus in his quotations from the Letters, Merrill eventually was able to compare the handwriting with the acknowledged script of Budaeus and to find that the two are identical.¹⁴ The Bodleian book, then, is not Aldus’s copy for the printer. It is Budaeus’s own collation from the Parisinus. Whether he examined the manuscript directly or used a copy made by Giocondo is doubtful; the note at the end of the Bodleian volume seems to favor the latter possibility. Budaeus does not by any means give a complete collation, but what he does give constitutes, in Merrill’s opinion, our best authority for any part of the lost Parisinus.¹⁵

The Morgan fragment possibly a part of the lost Parisinus

The script Perhaps we may now say the Bodleian volume has been hitherto our best authority. For a fragment of the ancient book, if my conjecture is right, is now, after various journeys, reposing in the Pierpont Morgan Library in New York City.

First of all, we are impressed with the script. It is an uncial of about the year 500 A.D.—certainly venerandae vetustatis. If Aldus had this same uncial codex at his disposal, we can understand his delight and pardon his slight exaggeration, for it is only slight. The essential truth of his statement remains: he had found a book of a different class from that of the ordinary manuscript—indeed diversis a nostris characteribus. Instead of thinking him arrant knave or fool enough to 41 bring down “antiquity” to the thirteenth century, we might charitably push back his definition of “nostri characteres” to include anything in minuscules; script “not our own” would be the majuscule hands in vogue before the Middle Ages. That is a position palaeographically defensible, seeing that the humanistic script is a lineal descendant of the Caroline variety. Furthermore, an uncial hand, though clear and regular as in our fragment, is harder to read than a glance at a page of it promises. This is due to the writing of words continuously. It takes practice, as Aldus says, to decipher such a script quickly and accurately. Moreover, the flesh sides of the leaves are faded.

Provenience and contents We next note that the fragment came to the Pierpont Morgan Library from Aldus’s country, where, as Dr. Lowe has amply shown, it was written; how it came into the possession of the Marquis Taccone would be interesting to know. But, like the Parisinus, the book to which our fragment belonged had not stayed in Italy always. It had made a trip to France—and was resting there in the fifteenth century, as is proved by the French note of that period on fol. 51^r. We may say “the book” and not merely “the present six leaves,” for the fragment begins with fol. 48, and the foliation is of the fifteenth century. The last page of our fragment is bright and clear, showing no signs of wear, as it would if no more had followed it;¹⁶ I will postpone the question of what probably did follow. Moreover, if the probatio pennae on fol.53^r is Carolingian,¹⁷ it would appear that the book had been in France at the beginning as well as at the end of the Middle Ages. Thus our manuscript may well have been one of those brought up from Italy by the emissaries of Charlemagne or their successors during the revival of learning in the eighth and ninth centuries. The outer history of our book, then, and the character of its script, comport with what we know of Aldus’s Parisinus.

The text closely related to that of Aldus But we must now subject our fragment to internal tests. If Aldus used the entire manuscript of which this is a part, his text must show a general conformity to that of the fragment. An examination of the appended collation will establish this fact beyond a doubt. The references are to Keil’s critical edition of 1870, but the readings are verified from Merrill’s apparatus. I will designate the fragment as Π, using P for Aldus’s Parisinus and a for his edition.

We may begin by excluding two probable misprints in Aldus, 64, 1 conturbernium and 65, 17 subeuertas. Then there are various spellings in which Aldus adheres to the fashion of his day, as sexcenties, millies, millia, tentarunt, caussas, autoritas, quanquam, syderum, hyeme, coena, ocium, hospicii, negociis, solatium, adulescet, 42 exoluit, Thuscos; there are other spellings which modern editors might not disdain, i.e., aerarii and illustri, and some that they have accepted, namely apponitur, existat, impleturus, implorantes, obtulissem, balinei, caret (not karet), caritas (not karitas).¹⁸

A study of our collation will also show some forty cases of correction in Π by either the scribe himself or a second and possibly a third ancient hand. Here Aldus, if he read the pages of our fragment and read them with care, might have seen warrant for following either the original text or the emended form, as he preferred. The most important cases are:61, 14 sera] Πa serua Π² 61, 21 considit] Π considet Π²a. The original reading of Π is clearly considit. The second i has been altered to a capital e, which of course is not the proper form for uncial. 62, 5 residit] Π residet a. Here Π is not corrected, but Aldus may have thought that the preceding case of considet (m. 2) supported what he supposed the better form residet. 63, 11 posset] a possit (in posset m. 1?) Π. Again the corrected e is capital, not uncial, but Aldus would have had no hesitation in adopting the reading of the second hand. 64, 2 modica vel etiam] a modica est etiam (corr. m. 2) Π. 64, 28 excurrissem accepto, ut praefectus aerari, commeatu] a. Here Π omitted accepto ut praefectus aerari,—evidently a line of the manuscript that he was copying, for there are no similar endings to account otherwise for the omission. 66, 2 dissentientis] a ex dissitientis m. 1 (?) Π.

There are also a few careless errors of the first hand, uncorrected, in Π, which Aldus himself might easily have corrected or have found the right reading already in the early editions. 62, 23 conteror quorum] a conteror qui horum Π B F 63, 28 si] a sibi Π 64, 24 conprobasse] comprouasse Π.

In view of these certain errors of the first hand of Π, most of them corrected but a few not, Aldus may have felt justified in abiding by one of the early editions in the following three cases, where Π might well have seemed to him wrong; in 43 one of them (64,3) modern editors agree with him: 62, 20 aurium oculorum vigor] Π aurium oculorumque uigor a 64, 3 proferenda] a conferanda Π 65, 11 et alii] Π etiam alii a.

There is only one case of possible emendation to note: 64, 29 questuri] Π quaesturi MVa. Aldus’s reading, as I learn from Professor Merrill, is in the anonymous edition ascribed to Roscius (Venice, 1492?), but not in any of the editions cited by Keil. This may be a conscious emendation, but it is just as possibly an error of hearing made by either Aldus or his compositor in repeating the word to himself as he wrote or set up the passage. Once in the text, quaesturi gives no offense, and is not corrected by Aldus in his edition of 1518. An apparently more certain effort at emendation is reported by Keil on 62, 13, where Aldus is said to differ from all the manuscripts and the editions in reading agere for facere. So he does in his second edition; but here he has facere with everybody else. The changes in the second edition are few and are largely confined to the correction of obvious misprints. There is no point in substituting agere for facere. I should attribute this innovation to a careless compositor, who tried to memorize too large a bit of text, rather than to an emending editor. At all events, it has no bearing on our immediate concern.

The striking similarity, therefore, between Aldus’s text and that of our fragment confirms our surmise that the latter may be a part of that ancient manuscript which he professes to have used in his edition. Whatever his procedure may have been, he has produced a text that differs from Π only in certain spellings, in the correction, with the help of existing editions, of three obvious errors of Π and of three of its readings that to Aldus might well have seemed erroneous, in two misprints, and in one reading which is possibly an emendation but which may just as well be another misprint. Thus the internal evidence of the text offers no contradiction of what the script and the history of the manuscript have suggested. I can not claim to have established an irrefutable conclusion, but the signs all point in one direction. I see enough evidence to warrant a working hypothesis, which we may use circumspectly as a clue, submit to further tests, and abandon in case these tests yield evidence with which it can not be reconciled.

Editorial methods of Aldus Further, if we are justified in our assumption that Aldus used the manuscript of which Π is a part, the fragment is instructive as to his editorial methods. If he proceeded elsewhere as carefully as here, he certainly did not perform his task with the high-handedness of the traditional humanistic editor; rather, he treated his ancient witness with respect, and abandoned it only when confronted with what seemed its obvious mistakes. I will revert to this matter at a later stage of the argument.

B UT, it will be asked, how do we know that Aldus used Π rather than some other manuscript that had a very similar text and that happened to have gone through the same travels? To answer this question we must examine the relation of Π to the other extant manuscripts in the light of what is known of the transmission of Pliny’s Letters in the Middle Ages. A convenient summary is given by Merrill on the basis of his abundant researches.¹⁹

Classes of the manuscripts Manuscripts of the Letters may be divided into three classes, distinguished by the number of books that each contains.

Class I, the ten-book family, consists of B (Bellovacensis or Riccardianus), now Ashburnhamensis, R 98 in the Laurentian Library in Florence, its former home, whence it had been diverted on an interesting pilgrimage by the noted book-thief Libri. This manuscript is attributed to the tenth century by Merrill, and by Chatelain in his description of the book. But Chatelain labels his facsimile page “Saec. IX.”²⁰ The latter seems the more probable date. The free use of a flat-topped a, along with the general appearance of the script, reminds me of the style in vogue at Fleury and its environs about the middle of the ninth century. A good specimen is accessible in a codex of St. Hilary on the Psalms (Vaticanus Reginensis 95), written at Micy between 846 and 859, of which a page is reproduced by Ehrle and Liebaert.²¹ F (Florentinus), the other important representative of this class, is also in the Laurentian Library (S. Marco 284). The date assigned to it seems also too late. It is apparently as early as the tenth century, and also has some of the characteristics of the script of Fleury; it is French work, at any rate. Keil’s suggestion²² that it may be the book mentioned as liber epistolarum Gaii Plinii in a tenth-century catalogue of the manuscripts at 45 Lorsch may be perfectly correct; though not written at Lorsch, it might have been presented to the monastery by that time.²³ These two manuscripts agree in containing, by the first hand, only Books I-V, vi (F having all and B only a part of the sixth letter). However, as the initial title in B is plini · secundi · epistularum · libri · decem, we may infer that some ancestor, if not the immediate ancestor, of B and F had all ten books.

In Class II the leading manuscript is another Laurentian codex (Mediceus XLVII 36), which contains Books I-IX, xxvi, 8. It was written in the ninth century, at Corvey, whence it was brought to Rome at the beginning of the sixteenth century. It is part of a volume that also once contained our only manuscript of the first part of the Annals of Tacitus.²⁴ The other chief manuscript of this class is V (Vaticanus Latinus 3864), which has Books I-IV. The script has been variously estimated. I am inclined to the opinion that the book was written somewhere near Tours, perhaps Fleury, in the earlier part of the ninth century.²⁵ If Ullman is right in seeing a reference to Pliny’s Letters in a notice in a mediaeval catalogue of Corbie,²⁶ it may be that the codex is a Corbeiensis. But it is also possible that a volume of the Letters at Corbie was twice copied, once at Corvey (M) and once in the neighborhood of Tours (V). At any rate, with the help of V, we may reach farther back than Corvey and Germany for the origin of this class. There are likewise two fragmentary texts, both of brief extent, Monacensis 14641 (olim Emmeramensis) saec. IX, and Leidensis Vossianus 98 saec. IX, the latter partly in Tironian notes. Merrill regards these as bearing “testimony to the existence of the nine-book text in the same geographical region,” namely Germany.²⁷ There they are to-day, in Germany and Holland, but where they were written is another affair. The Munich fragment is part of a composite 46 volume of which it occupies only a page or two. The script is continental, and may well be that of Regensburg, but it shows marked traces of insular influence, English rather than Irish in character. The work immediately preceding the fragment is in an insular hand, of the kind practised at various continental monasteries, such as Fulda; there are certain notes in the usual continental hand. Evidently the manuscript deserves consideration in the history of the struggle between the insular and the continental hands in Germany.²⁸ The script of the Leyden fragment, on the other hand, so far as I can judge from a photograph, looks very much like the mid-century Fleury variety with which I have associated the Bellovacensis; there can hardly be doubt, at any rate, that De Vries is correct in assigning it to France, where Voss obtained so many of his manuscripts.²⁹ Except, therefore, for M and the Munich fragment, there is no evidence furnished by the chief manuscripts which connects the tradition of the Letters with Germany. The insular clue afforded by the latter book deserves further attention, but I can not follow it here. The question of the Parisinus aside, B and F of Class I and V of Class II are sure signs that the propagation of the text started from one or more centres—Fleury and Corbie seem the most probable—in France.

The third class comprises manuscripts containing eight books, the eighth being omitted and the ninth called the eighth. Representatives of this class are all codices of the fifteenth century, though the class has a more ancient basis than that, namely a lost manuscript of Verona. This is best attested by D, a Dresden codex, while almost all other manuscripts of this class descend from a free recension made by Guarino and conflated with F; o, u, and x are the representatives of this recension (G) that are reported by Merrill. The relation of this third class to the second is exceedingly close; indeed, it may be merely a branch of it.³⁰

The early editions As is often the case, the leading manuscript authorities are only inadequately represented in the early editions. The Editio Princeps (p) of 1471 was based on 47 a manuscript of the Guarino recension. A Roman editor in 1474 added part of Book VIII, putting it at the end and calling it Book IX; he acquired this new material, along with various readings in the other books, from some manuscript of Class II that may have come down from the north. Three editors, called ς by Keil—Pomponius Laetus 1490, Beroaldus 1498, and Catanaeus 1506—took r as a basis; but Laetus had another and a better representative of the same type of text as that from which r had drawn, and he likewise made use of V. With the help of these new sources the ς editors polished away a large number of the gross blunders of p and r, and added a sometimes unnecessary brilliance of emendation. Avantius’s edition of part of Book X in 1502 was appropriated by Beroaldus in the same year and by Catanaeus in 1506; these latter editors had no new sources at their disposal. No wonder that the Parisinus seemed a godsend to Aldus. The only known ancient manuscripts whose readings had been utilized in the editions preceding his own were F and V, both incomplete representatives of Classes I and II. The manuscripts discovered by the Roman editor and Laetus were of great help at the time, but we have no certain evidence of their age. B and M were not accessible.³¹ Now, besides the transcript of Giocondo and his other six volumes, whatever these may have been, Aldus had the ancient codex itself with all ten books complete. Everybody admits that the Parisinus, as shown by the readings of Aldus, is clearly associated with the manuscripts of Class I. Its contents corroborate the evidence of the title in B, which indicates descent from some codex containing ten books.

Π a member of Class I Now nothing is plainer than that Π is a member of Class I, as it agrees with BF in the following errors, or what are regarded by Keil as errors. I consider the text of the Letters and not their superscriptions. 60, 15 duplicia] MVD duplicata ΠBFGa; 61, 12 confusa adhuc] MV adhuc confusa ΠBFGa; 62, 6 doctissime] MV doctissima ΠBFDa et doctissima G; 62, 16 nec adficitur] MVD et adficitur ΠBFGa; 62, 23 quorum] MVDGa qui horum ΠBF; 63, 22 teque et] MVDG teque ΠBFa; 64, 3 proferenda] Doxa conferenda BFu conferanda Π (MV lack an extensive passage here); 65, 11 alii quidam minores sed tamen numeri] DG alii quidam minores sed tam innumeri MV alii quidem minoris sed tamen numeri ΠBFa; 65, 12 voluntariis accusationibus] M (uoluntaris) D voluntariis om. V accusationibus uoluntariis ΠBFGa; 65, 15 superiore] MVD priore ΠBFGa; 65, 24 iam] MVDG om. ΠBFa.

Tastes differ, and not all these eleven readings of Class I may be errors. Kukula, in the most recent Teubner edition (1912), accepts three of them (60, 15; 62, 6; 65, 15), and Merrill, in his forthcoming edition, five (60, 15; 61, 12; 62, 6; 48 65, 12; 65, 15). Personally I could be reconciled to them all with the exception of the very two which Aldus could not admit—62, 23 and 64, 3; in both places he had the early editions to fall back on. However, I should concur with Merrill and Kukula in preferring the reading of the other classes in 62, 16 and 65, 24. In 65, 11 I would emend to alii quidam minoris sed tamen numeri; if this is the right reading, ΠBF agree in the easy error of quidem for quidam, and MVD in another easy error, minores for minoris—the parent manuscript of MV further changed tamen numeri to tam innumeri. Whatever the final judgment, here are five cases in which all recent editors would attribute error to Class I; in the remaining six cases the manuscripts of Class I either agree in error or avoid the error of Class II—surely, then, Π is not of the latter class. There are six other significant errors of MV in the whole passage, no one of which appears in Π: 61, 15 si non] sint MV; 62, 6 mira illis] mirabilis MV; 62, 11 lotus] illic MV; cibum] cibos MV; 62, 25 fuit—64, 12 potes] om. MV; 66, 12 amatus] est amatus MV. Once the first hand in Π agrees with V in an error easily committed independently: 61, 12 ordinata] ordinata, di ss. m. 2 Π ornata V.

Π, then, and MV have descended from the archetype by different routes. With Class III, the Verona branch of Class II, Π clearly has no close association.

But the evidence for allying Π with B and F, the manuscripts of Class I, is by no means exhausted. In 61, 14, BFux have the erroneous emendation, which Budaeus includes among his variants, of serua for sera. A glance at Π shows its apparent origin. The first hand has sera correctly; the second hand writes u above the line.³² If the second hand is solely responsible for the attempt at improvement here, and is not reproducing a variant in the parent manuscript of Π, then BF must descend directly from Π. The following instances point in the same direction: 61, 21 considit] considet BF. Π has considit by the first hand, the second hand changing the second i to a capital e.³³ In 65, 5, however, residit is not thus changed in Π, and perhaps for this very reason is retained by the careful scribe of B; F, which has a slight tendency to emend, has, with G, residet. 63, 9 praestat amat me] praestatam ad me B. Here the letters of the scriptura continua in Π are faded and blurred; the error of B would therefore be peculiarly easy if this manuscript derived directly from Π. If one ask whether the page were as faded in the ninth century as now, Dr. Lowe has already answered this question; the flesh side of the parchment might well have lost a portion of its ink considerably before the Carolingian period.³⁴ In any case, the error of praestatam ad me seems natural enough to one who reads the line for the first time in Π. B did not, as we shall see, copy directly from Π; a copy intervened, in which the error was made and then, I should infer, corrected above the line, whence F drew the right reading, B taking the original but incorrect text.

49 There are cases in plenty elsewhere in the Letters to show that B is not many removes from the scriptura continua of some majuscule hand. In the section included in Π, apart from the general tightness of the writing, which led to the later insertion of strokes between many of the words,³⁵ we note these special indications of a parent manuscript in majuscules. In 61, 10 me autem], B started to write mea and then corrected it. 64, 19 praeceptori a quo] praeceptoria quo B, (m. 1) F. If B or its parent manuscript copied Π directly, the mistake would be especially easy, for praeceptoria ends the line in Π. 64, 25 integra re]. After integra, a letter is erased in B; the copyist, it would seem, first mistook integra re for one word.

Other instances showing a close connection between B and Π are as follows: 62, 23 unice] Π has by the first hand inuice, the second hand writing u above i, and a vertical stroke above u. In BF, uince, the reading of the first hand, is changed by the second to unice; this second hand, Professor Merrill informs me, seems to be that of a writer in the same scriptorium as the first. The error in BF might, of course, be due to copying an original in minuscules, but it might also be due to the curious state of affairs in Π. 65, 24 fungerer]. In Π the final r is written, somewhat indistinctly, above the line. B has fungerer corrected by the second hand from fungeret (?), which may be due to a misunderstanding of Π. 66, 2 avunculi] auonculi Π (o in ras.) B. This form might perhaps be read; F has emended it out, and no other manuscript has it. 65, 7 desino, inquam, patres conscripti, putare]. Here the relation of BF to Π seems particularly close. Π, like MVDoxa, has the abbreviation p.c. On a clearly written page, the error of reputare (BF) for p.c. putare is not a specially likely one to make. But in the blur at the bottom of fol. 52^v, a page on the flesh side of the parchment, the combination might readily be mistaken for reputare.

Another curious bit of testimony appears at the beginning of the third book. The scribe of B³⁶ wrote the words nescio—apud in rustic capitals, occupying therewith the first line and about a third of the second. This is not effective calligraphy. It would appear that he is reproducing, as is his habit, exactly what he found in his original. That original might have had one full line, or two lines, of majuscules, perhaps, following pretty closely the lines in Π, which has the same amount of text, plus the first three letters of spurinnam, in the first two lines. If B had Π before him, there is nothing to explain his most unusual procedure. His original, therefore, is not Π but an intervening copy, which he is transcribing with an utter indifference to aesthetic effect and with a laudable, if painful, desire for accuracy. This trait, obvious in B’s work throughout, is perhaps nowhere more strikingly exhibited than here.

Π the direct ancestor of BF with probably a copy intervening If Π is the direct ancestor of BF, these manuscripts should contain no good readings not found in Π, unless their writers could arrive at such readings by easy emendation or unless there is contamination with some other source. From what we know of the text of BF in general, the latter supposition may at once be ruled out. There are but three cases to consider, two of which may be readily disposed of: 64, 3 proferenda] conferenda BF conferanda Π; 64, 4 conprobasse] (comp.) BF comprouasse Π. These are simple slips, which a scribe might almost unconsciously correct as he wrote. The remaining error (63, 28 sibi to si) is not difficult to emend when one considers the entire sentence: quibus omnibus ita demum similis adolescet, si imbutus honestis artibus fuerit, quas, etc. It is less probable, however, that B with Π before him should correct it as he wrote than, as we have already surmised, that a minuscule copy intervened between Π and B, in which the letters bi were deleted by some careful reviser. Two other passages tend to confirm this assumption of an intermediate copy. In 65, 6 (tum optime libertati venia obsequio praeparatur), B has optimae, a false alteration induced perhaps by the following libertati. In Π, optime stands at the end of the line. The scribe of B, had he not found libertati immediately adjacent, would not so readily be tempted to emend; still, we should not make too much of this instance, as B has a rather pronounced tendency to write ae for e. A more certain case is 66, 7 fungar indicis] fungarindicis ex fungari dicis B; here the error is easier to derive from an original in minuscules in which in was abbreviated with a stroke above the i. There is abundant evidence elsewhere in the Letters that the immediate ancestor of BF was written in minuscules; I need not elaborate this point. Our present consideration is that apart from the three instances of simple emendation just discussed, there is no good reading of B or F in the portion of text contained in Π that may not be found, by either the first or the second hand, in Π.³⁷

We may now examine a most important bit of testimony to the close connection existing between BF and Π. B alone of all manuscripts hitherto known is provided with indices of the Letters, one for each book, which give the names of the correspondents and the opening words of each letter. Now Π, by good luck, preserves the end of Book II, the beginning of Book III, and between them the index for Book III. Dr. F. E. Robbins, in a careful article on B and F, and one 51 on the tables of contents in B,³⁸ concluded that P did not contain the indices which are preserved in B, and that these were compiled in some ancestor of B, perhaps in the eighth century. Here they are, in the Morgan fragment, which takes us back two centuries farther into the past. A comparison of the index in Π shows indubitably a close kinship with B. A glance at plates XIII and XIV indicates, first of all, that the copy B, here as in the text of the Letters, is not many removes from scriptura continua. Moreover, the lists are drawn up on the same principle; the nomen and cognomen but not the praenomen of the correspondent being given, and exactly the same amount of text quoted at the beginning of each letter. The incipit of III, xvi (ad nepotem—adnotasse uideor fatadictaq·) is an addition in Π, and the lemma is longer than usual, as though the original title had been omitted in the manuscript which Π was copying and the corrector of Π had substituted a title of his own making.³⁹ It reappears in B, with the easy emendation of facta from fata. The only other case in the indices of a right reading in B that is not in Π is in the title of III, viii: ad sueton tranque Π Adsu&on tranqui. B. In both these instances the scribe of B needed no external help in correcting the simple error. Far more significant is the coincidence of B and Π in very curious mistakes, as the address of III, iii (ad caerelliae hispullae for ad corelliam hispullam) and the lemma of III, viii (facis adprocetera for facis pro cetera). ΠBF agree in omitting suae (III, iii) and suo (III, iv), but in retaining the pronominal adjectives in the other addresses preserved in Π. The same unusual suspensions occur in Π and B, as ad sueton tranque (tranqui B); ad uestric spurinn·; ad silium procul.⁴⁰ In the first of these cases, the parent of Π evidently had tranq·, which Π falsely enlarges to tranque; this form and not tranq· is the basis of B’s correction—a semi-successful correction—tranqui. This, then, is another sign that B depends directly on Π. Further, B omits one symbol of abbreviation which Π has (possum iam perscri{-b}), the lemma of the ninth letter), and in the lemma of the tenth neither manuscript preserves the symbol (composuisse me quaed). In the first of these cases, it will be observed, B has a very long i in perscrib.⁴¹ This long i is not a feature of the script of B, nor is there any provocation for it in the way in which the word is written in Π. This detail, therefore, may be added to the indications that a copy in minuscules intervened between B and Π; the curious i, faithfully reproduced, as usual, by B, may have occurred in such a copy.

These details prove an intimate relation between Π and BF, and fit the supposition that B and F are direct descendants of Π. This may be strengthened 52 by another consideration. If Π and B independently copy the same source, they inevitably make independent errors, however careful their work. Π should contain, then, a certain number of errors not in B. As we have found only three such cases in 12 pages, or 324 lines, and as in all these three the right reading in B could readily have been due to emendation on the part of the scribe of B or of a copy between Π and B, we have acquired negative evidence of an impressive kind. It is distinctly harder to believe that the two texts derive independently from a common source. Show us the significant errors of Π not in B, and we will accept the existence of that common source; otherwise the appropriate supposition is that B descends directly from its elder relative Π. It is not necessary to prove by an examination of readings that Π is not copied from B; the dates of the two scripts settle that matter at the start. Supposing, however, for the moment, that Π and B were of the same age, we could readily prove that the former is not copied from the latter. For B contains a significant collection of errors which are not present in Π. Six slight mistakes were made by the first hand and corrected by it, three more were corrected by the second hand, and twelve were left uncorrected. Some of these are trivial slips that a scribe copying B might emend on his own initiative, or perhaps by a lucky mistake. Such are 64, 26 iudicium] indicium B; 64, 29 Caecili] caecilii B; 65, 13 neglegere] neglere B. But intelligent pondering must precede the emendation of praeceptoria quo into praeceptori a quo (64, 19), of beaticis into Baeticis (65, 15), and of optimae into optime (65, 26), while it would take a Madvig to remedy the corruptions in 63, 9 (praestatam ad me) and 65,7 (reputare into patres conscripti putare). These are the sort of errors which if found in Π would furnish incontrovertible proof that a manuscript not containing them was independent of Π; but there is no such evidence of independence in the case of B. Our case is strengthened by the consideration that various of the errors in B may well be traced to idiosyncrasies of Π, not merely to its scriptura continua, a source of misunderstanding that any majuscule would present, but to the fading of the writing on the flesh side of the pages in Π, and to the possibility that some of the corrections of the second hand may be the private inventions of that hand.⁴² We are hampered, of course, by the comparatively small amount of matter in Π, nor are we absolutely certain that this is characteristic of the entire manuscript of which it was once a part. But my reasoning is correct, I believe, for the material at our disposal.

The probable stemma Our tentative stemma thus far, then, is No. 1 below, not No. 2 and not No. 3.

Robbins put P in the position of Π in this last stemma, but on the assumption that it did not contain the indices. That is not true of Π.

Further consideration of the external history of P, Π, and B Still further evidence is supplied by the external history of our manuscripts. B was at Beauvais at the end of the twelfth or the beginning of the thirteenth century, as we have seen.⁴³ Whatever the uncertainties as to its origin, any palaeographer would agree that it could hardly have been written before the middle of the ninth century or after the middle of the tenth. It was undoubtedly produced in France, as was F, its sister manuscript. The presumption is that Π¹, the copy intervening between Π and B, was also French, and that Π was in France when the copy was made from it. Merrill, for what reason I fail to see, suggested that the original of BF might be “Lombardic,” written in North Italy.⁴⁴ An extraneous origin of this sort must be proved from the character of the errors, such as spellings and the false resolution of abbreviations, made by BF. If no such signs can be adduced, it is natural to suppose that Π¹ was of the same nationality and general tendencies as its copies B and F. This consideration helps out the possible evidence furnished by the scribbling in a hand of the Carolingian variety on fol. 53^v;⁴⁵ we may now be more confident that it is French rather than Italian. But whatever the history of our book in the early Middle Ages, in the fifteenth century it was surely near Meaux, which is not far from Paris—about as far to the east as Beauvais is to the north. Now, granted for a moment that the last of our stemmata is correct, X, from which Π and B descend, being earlier than Π, must have been a manuscript in majuscules, written in Italy, since that is unquestionably the provenience of Π. There were, then, by this supposition, two ancient majuscule manuscripts of the Letters, most closely related in text—veritable twins, indeed—that travelled from Italy to France. One (X¹) had arrived in the early Middle Ages and is the parent of 54 B and F; the other (Π) was probably there in the early Middle Ages, and surely was there in the fifteenth century. We can not deny this possibility, but, on the principle melius est per unum fieri quam per plura, we must not adopt it unless driven to it. The history of the transmission of Classical texts in the Carolingian period is against such a supposition.⁴⁶ Not many books of the age and quality of Π were floating about in France in the ninth century. There is nothing in the evidence presented by Π and B that drives us to assume the presence of two such codices. There is nothing in this evidence that does not fit the simpler supposition that BF descend directly from Π. The burden of proof would appear to rest on those who assert the contrary. Π, therefore, if the ancestor of B, contained at least as much as we find today in B. Some ancestor of B had all ten books. Aldus, whose text is closely related to BF, got all ten books from a very ancient manuscript that came down from Paris. Our simpler stemma indicates the presence of one rather than more than one such manuscript in the vicinity of Paris in the ninth or the tenth century and again in the fifteenth. This line of argument, which presents not a mathematically absolute demonstration but at least a highly probable concatenation of facts and deductions, warrants the assumption, to be used at any rate as a working hypothesis, that Π is a fragment of the lost Parisinus which contained all the books of Pliny’s Letters.

Our stemma, then, becomes,

P (the whole manuscript), of which Π is a part.

Evidence from the portions of BF outside the text of Π We may corroborate this reasoning by evidence drawn from the portions of BF outside the text of Π. We note, above all, a number of omissions in BF that indicate the length of line in some manuscript from which they descend. This length of line is precisely what we find in Π. Our fragment has lines containing from 23 to 33 letters, very rarely 23, 24, or 33, and most frequently from 27 to 30, the average being 28.4. These figures tally closely with those given by Professor A. C. Clark⁴⁷ for the Vindobonensis of Livy, a codex not far removed in 55 date from Π. Supposing that Π is a typical section of P—and after Professor Clark’s studies⁴⁸ we may more confidently assume that it is—P had the same length of line. The important cases of omission are as follows:

32, 19 atque etiam invisus virtutibus fuerat evasit, reliquit incolumen optimum atque] etiam—atque om. BF. P would have the abbreviation for bus in virtutibus and for que in atque. There would thus be in all 61 letters and dots, or two lines, arranged about as follows:

The scribe could easily catch at the second ATQ· after writing the first. It will be at once objected that the repeated ATQ· might have occasioned the mistake, whatever the length of the line. Thus in 82, 2 (aegrotabat Caecina Paetus, maritus eius, aegrotabat] Caecina—aegrotabat om. BF), the omitted portion comprises 34 letters—a bit too long, perhaps, for a line of P. The following instances, however, can not be thus disposed of.

94, 10 alia quamquam dignitate propemodum paria] quamquam—paria (32 letters) om. BF. Cetera and paria, to be sure, offer a mild case of homoioteleuta, but not powerful enough to occasion an omission unless the words happened to stand at the ends of lines, as they might well have done in P. As the line occurs near the beginning of a letter, we may verify our conjecture by plotting the opening lines. The address, as in Π, would occupy a line. Then, allowing for contractions in rebus (18) and quoque (19) and reading cum (Class I) for quod (18), cetera (Class I) for alia (20), we can arrange the 236 letters in 8 lines, with an average of 29.5 letters in a line.

123, 10 sentiebant. interrogati a Nepote praetore quem docuissent, responderunt quem prius: interrogati an tunc gratis adfuisset, responderunt sex milibus] interrogati a Nepote—docuissent responderunt om. BF. Here are two good chances for omissions due to similar endings, as interrogati and responderunt are both repeated, but neither chance is taken by BF. Instead, a far less striking case (sentiebant—responderunt) leads to the omission. The arrangement in P might be

Here the dangerous words interrogati and responderunt are in safe places.

We may add two omissions of F in passages now missing altogether in B. 69, 28 quod minorem ex liberis duobus amisit sed maiorem] minorem—sed om. F. Here again an omission is imminent from the similar endings minorem—maiorem; that made by F (29 letters and one dot) seems to be that of a line of P where the arrangement would be:

There may have been a copy (P²) intervening between P¹ and F, but doubtless neither that nor P¹ itself had lines so short as those in P; the error of F, therefore, may be most naturally ascribed to P¹, who omitted a line of P.

130, 16 percolui. in summa (cur enim non aperiam tibi vel iudicium meum vel errorem?) primum ego] in summa—primum (59 letters) om. F. As there are no homoioteleuta here at all, we surely are concerned with the omission of a line or lines. Perhaps 59 letters would make up a line in P¹ or P². Perhaps two lines of P were dropped.

Similarly we may note two omissions in B, though not in F, which may be due originally to the error of P¹ in copying P.

68, 5 electorumque commentarios centum sexaginta mihi reliquit, opisthographos] -torumque—opisthographos om. B. Allowing the abbreviation of que, we have 59 letters and one dot here. The omitted words are written by the first hand of B at the foot of the page. Of course the omission may correspond to a line of P¹ dropped by B in copying, but it is equally possible that P¹ committed the error and corrected it by the marginal supplement, F noting the correction in time to include the omitted words in his text, B copying them in the margin as he found them in P¹.

87, 12 tacitus suffragiis impudentia inrepat. nam quoto cuique eadem honestatis] suffragiis—honestatis om. m. 1, add. in mg. m. 2 B (54 letters, with que abbreviated). This may be like the preceding, except that the correction was done not by the original scribe of B, but by a scribe in the same monastery. The presence of homoioteleuta, we must admit, adds an element of uncertainty.

So, of the passages here brought forward, 94, 20; 123, 10 and 69, 28 are best explained by supposing that B and F descend from a manuscript that like Π had 57 from 24 to 32 letters in a line, while 32, 19 and 130, 16 fit this supposition as well as they do any other.

One orthographic peculiarity is perhaps worth noting: we saw that B did not agree with Π in the spellings karet and karitas.⁴⁹ We do, however, find karitate elsewhere in B (109, 8), and the curious reading Kl∴facere, mg. calfacere, for calfacere (56, 12). This is an additional bit of evidence for supposing that a copy (P¹) intervened between P and B; P had the spelling Karitas consistently, P¹ altered it to the usual form, and B reproduced the corrections in P¹, failing to take them all, unless, as may well be, P¹ had failed to correct all the cases.

Thus the evidence contained in the portion of BF outside the text of Π corroborates our working hypothesis deduced from the fragment itself. We have found nothing yet to overthrow our surmise that a bit of the ancient Parisinus is veritably in the city of New York.

Aldus’s methods; his basic text W E may now return to Aldus and imagine, if we can, his method of critical procedure. Finding his agreement with Π so close, even in what editors before and after him have regarded as errors, I am disposed to think that he studied his Parisinus with care and followed its authority respectfully. Finding that his seemingly extravagant statements about the antiquity of his book are essentially true, I am disposed to put more confidence in Aldus than editors have granted him thus far. I should suppose that, working in the most convenient way, he turned over to his compositor, not a fresh copy of P, but the pages of some edition corrected from P—which Aldus surely tells us that he used—and from whatever other sources he consulted. It may be beyond our powers to discover the precise edition that he thus employed. It does not at first thought seem likely that he would select the Princeps, which does not include the eighth book at all, and contains errors that later were weeded out. In the portion of text included in Π, P has thirty-two readings which Aldus avoids. In most of these cases p commits an error, sometimes a ridiculous error, like offam for officia (62, 25); the manuscript on which p was based apparently made free use of abbreviations. Keil’s damning estimate of r⁵⁰ is amply borne out in this section of the text; Aldus differs from r in sixty-five cases, most of these being errors in r. He agrees with ς in all but twenty-six readings.⁵¹ Aldus would have had fewest changes to make, then, if his basic text was ς. This is apparently the view of Keil,⁵² who would agree at any rate that Aldus made special use of the ς editions and who also declares that p is the fundamentum of r as r is of the edition of Pomponius Laetus.⁵³

It would certainly be natural for Aldus to start with his immediate predecessors, as they had started with theirs. The matter ought to be cleared up, if possible, for in order to determine what Aldus found in P we must know whether he took some text as a point of departure and, if so, what that text was. But the task should be undertaken by some one to whom the early editions are accessible. Keil’s report of them, intentionally incomplete,⁵⁴ is sufficient, he declares,⁵⁵ “ad fidem Aldinae editionis constituendam,” but, as I have found by comparing our photographs of the edition of Beroaldus in the present section, Keil has not collated minutely or accurately enough to encourage us to undertake, on the basis of his apparatus, an elaborate study of Aldus’s relation to the editions preceding his own.

59 The variants of Budaeus in the Bodleian volume We may now test Aldus by the evidence of the Bodleian volume with its variants in the hand of Budaeus. For the section included in Π, their number is disappointingly small. The only additions by Budaeus (= i) to the text of Beroaldus are: 61, 14 sera] MVDoa, (m. 1) Π serua BFuxi, (m. 2) Π; 62, 4 ambulat] i cum plerisque ambulabat r Ber. (ab del.) M; 62, 25 quoque] i cum ceteris p̷ouq (ue) Ber.; 64, 23 Quamvis] q Vmuis Ber. corr. i.

This is all. Budaeus, who, according to Merrill, had the Parisinus at his disposal, has corrected two obvious misprints, made an inevitable change in the tense of a verb—with or without the help of the ancient book—and introduced from that book one unfortunate reading which we find in the second hand of Π.

There is one feature of Budaeus’s marginal jottings that at once arouses the curiosity of the textual critic, namely, the frequent appearance of the obelus and the obelus cum puncto. These signs as used by Probus⁵⁶ would denote respectively a surely spurious and a possibly spurious line or portion of text. But such was not the usage of Budaeus; he employed the obelus merely to call attention to something that interested him. Thus at the end of the first letter of Book III we find a doubly pointed obelus opposite an interesting passage, the text of which shows no variants or editorial questionings. Budaeus appears to have expressed his grades of interest rather elaborately—at least I can discover no other purpose for the different signs employed. The simple obelus apparently denotes interest, the pointed obelus great interest, the doubly pointed obelus intense interest, and the pointing finger of a carefully drawn hand burning interest. He also adds catchwords. Thus on the first letter he calls attention successively⁵⁷ to Ambulatio, Gestatio, Hora balnei, pilae ludus, Coena, and Comoedi. The purpose of the doubly pointed obelus is plainly indicated here, as it accompanies two of these catchwords. Just so in the margin opposite 65, 17, a pointing finger is accompanied by the remark, “Beneficia beneficiis aliis cumulanda,” while 227, 5 is decorated with the moral ejaculation, “o hominem in diuitiis miserum.” Incidentally, it is obvious that the Morgan fragment was once perused by some thoughtful reader, who marked with lines or brackets passages of special interest to him. For example, the account of how Spurinna spent his day⁵⁸ is so marked. This passage likewise called forth various marginal notes from Budaeus,⁵⁹ and other coincidences exist between the markings in Π and the marginalia in the Bodleian volume. But there is not enough evidence of this sort to warrant the suggestion that Budaeus himself added the marks in Π.

Aldus and Budaeus compared It is of some importance to consider what Budaeus might have done to the text of Beroaldus had he treated it to a systematic collation with the Parisinus. Our fragment allows us to test Budaeus; for even if it be not the Parisinus itself, 60 its readings with the help of B, F, and Aldus show what was in that ancient book. I have enumerated above⁶⁰ eleven readings of ΠBF which are called errors by Keil, but of which nine were accepted by Aldus and five by the latest editor, Professor Merrill. In two of these (62, 33 and 64, 3), Budaeus, like Aldus, wisely does not harbor an obvious error of P. In two more (62, 16 and 65, 12), Beroaldus already has the reading of P. Of the remaining seven, however, all of which Aldus adopted, there is no trace in Budaeus. There are also nineteen cases of obvious error in the ς editions, which Aldus corrected but Budaeus did not touch. I give the complete apparatus⁶¹ for these twenty-six places, as they will illustrate the radical difference between Aldus and Budaeus in their use of the Parisinus.

Here is sufficient material for a test. Aldus, it will be observed, whether or not he started with some special edition, refuses to follow the latest and best texts of his day (i.e., ς) in these twenty-six readings. In one sure case (60, 15) and eleven possible⁶² cases (61, 18; 62, 26; 63, 5, 12, 15, 17 bis, 23 bis; 64, 2, 5), his reading agrees with the Princeps. In four sure cases (63, 4, 22; 65, 15; 66, 9) and one possible one (63, 9), he agrees with the Roman edition; in two sure (61, 12; 66, 11) and three possible (63, 2; 66, 7, 12) cases, with both p and r. Once he breaks away from all editions reported by Keil and agrees with D (62, 6). At the same time, all these readings are attested by ΠFB and hence were presumably in the Parisinus. In two cases (65, 11, 24), we know of no source other than P that could have furnished him his reading. Further, in the superscription of the third letter of Book III (63, 20), he might have taken a hint from Catanaeus, who was the first to depart from the reading corneliae, universally accepted before him, but again it is only P that could give him the correct spelling corelliae.⁶³

If all the above readings, then, were in the Parisinus, how did Aldus arrive at them? Did he fish round, now in the Princeps, now in the Roman edition, despite the repellent errors that those texts contained,⁶⁴ and extract with felicitous accuracy excellent readings that coincided with those of the Parisinus, or did he draw them straight from that source itself? The crucial cases are 65, 11 and 24. As he must have gone to the Parisinus for these readings, he presumably found the others there, too. Moreover, he did not get his new variants by a merely sporadic consultation of the ancient book when he was dissatisfied with the accepted text of his day, for in the two crucial cases and many of the others, too, that text makes sense; some of the readings, indeed, are accepted by modern editors as correct.⁶⁵ Aldus was collating. He carefully noted minutiae, such as the omission of et and iam, and accepted what he found, unless the ancient text seemed to him indisputably wrong. He gave it the benefit of the doubt even when it may be wrong. This is the method of a scrupulous editor who cherishes a proper veneration for his oldest and best authority.

Budaeus, on the other hand, is not an editor. He is a vastly interested reader 62 of Pliny, frequently commenting on the subject-matter or calling attention to it by marginal signs. As for the text, he generally finds Beroaldus good enough. He corrects misprints, makes a conjecture now and then, or adopts one of Catanaeus, and, besides supplementing the missing portions with transcripts made for him from the Parisinus, inserts numerous variants, some of which indubitably come from that manuscript.⁶⁶ In the present section, occupying 251 lines in Π, there is only one reading of the Parisinus—a false reading, it happens—that seems to Budaeus worth recording. Compared with what Aldus gleaned from Π, Budaeus’s extracts are insignificant. It is remarkable, for instance, that on a passage (65, 11) which, as the appended obelus shows, he must have read with attention, he has not added the very different reading of the Parisinus. Either, then, Budaeus did not consult the Parisinus with care, or he did not think the great majority of its readings preferable to the text of Beroaldus, or, as I think may well have been the case, he had neither the manuscript itself nor an entire copy of it accessible at the time when he added his variants in his combined edition of Beroaldus and Avantius.⁶⁷

But I do not mean to present here a final estimate of Budaeus; for that, I hope, we may look to Professor Merrill. Nor do I particularly blame Budaeus for not constructing a new text from the wealth of material disclosed in the Parisinus. His interests lay elsewhere; suos quoique mos. What I mean to say, and to say with some conviction, is that for the portion of text included in our fragment, the evidence of that fragment, coupled with that of B and F, shows that as a witness to the ancient manuscript Aldus is overwhelmingly superior to either Budaeus or any of the ancient editors.

Our examination of the Morgan fragment, therefore, leads to what I deem a highly probable conclusion. We could perhaps hope for absolute proof in a matter of this kind only if another page of the same manuscript should appear, bearing a note in the hand of Aldus Manutius to the effect that he had used the codex for his edition of 1508. Failing that, we can at least point out that all the data accessible comport with the hypothesis that the Morgan fragment was a part of this very codex. We have set our hypothesis running a lengthy gauntlet of facts, and none has tripped it yet. We have also seen that Π is most intimately connected with manuscripts BF of Class I, and indeed seems to be a part of the very manuscript whence they are descended. Finally, a careful comparison of Aldus’s text 63 with Π shows him, for this much of the Letters at least, to be a scrupulous and conscientious editor. His method is to follow Π throughout, save when, confronted by its obvious blunders, he has recourse to the editions of his day.

The latest criticism of Aldus Since the publication of Otto’s article in 1886,⁶⁸ in which the author defended the F branch against that of MV, to which, as the elder representative of the tradition, Keil had not unnaturally deferred, critical procedure has gradually shifted its centre. The reappearance of B greatly helped, as it corroborates the testimony of F. B and F head the list of the manuscripts used by Kukula in his edition of 1912,⁶⁹ and B and F with Aldus’s Parisinus make up Class I, not Class II, in Merrill’s grouping of the manuscripts. Obviously, the value of Class I mounts higher still now that we have evidence in the Morgan fragment of its existence in the early sixth century. This fact helps us to decide the question of glosses in our text. We are more than ever disposed to attribute not to BF but to what has now become the younger branch of the tradition, Class II, the tendency to interpolate explanatory glosses. The changed attitude towards the BF branch has naturally resulted in a gradual transformation of the text. We have seen in the portion included in Π that of the eleven readings which Keil regarded as errors of the F branch, three are accepted by Kukula and five by Merrill.⁷⁰

Since Class I has thus appreciated in value, we should expect that Aldus’s stock would also take an upward turn. In Aldus’s lifetime, curiously, he was criticized for excessive conservatism. His rival Catanaeus finds his chief quality supina ignorantia and adds:⁷¹

“Verum enim uero non satis est recuperare venerandae vetustatis exemplaria, nisi etiam simul adsit acre emendatoris iudicium: quoniam et veteres librarii in voluminibus describendis saepissime falsi sunt, et Plinius ipse scripta sua se viuo deprauari in quadam epistola demonstrauerit.”

Nowadays, however, editors hesitate to accept an unsupported reading of Aldus as that of the Parisinus, since they believe that he abounds in those very conjectures of which Catanaeus felt the lack. The attitude of the expert best qualified to judge is still one of suspicion towards Aldus. In his most recent article,⁷² Professor Merrill declares that Keil’s remarks⁷³ on the procedure of Aldus in the part of Book X already edited by Avantius, Beroaldus, and Catanaeus might safely have been extended to cover the work of Aldus on the entire body of the Letters. He proceeds to subject Aldus to a new test, the material for which we owe to Merrill’s own researches. He compares with Aldus’s text the manuscript parts of the Bodleian 64 volume, which are apparently transcripts from the Parisinus (= I);⁷⁴ in them Budaeus with his own hand (= i) has corrected on the authority of the Parisinus itself, according to Merrill, the errors of his transcriber. In a few instances, Merrill allows, Budaeus has substituted conjectures of his own. This material, obviously, offers a valuable criterion of Aldus’s methods as an editor. There is a further criterion in the shape of Codex M, not utilized till after Aldus’s edition. As this manuscript represents Class II, concurrences between M and Ii against a make it tolerably certain that Aldus himself and no higher authority is responsible for such readings. On this basis, Merrill cites twenty-five readings in the added part of Book VIII (viii, 3 quas obvias—xviii, II amplissimos hortos) and nineteen readings in the added part of Book X (letters iv-xli), which represent examples “wherein Aldus abandons indubitably satisfactory readings of his only and much belauded manuscript in favor of conjectures of his own.”⁷⁵ Letter IX xvi, a very short affair, added by Budaeus in the margin, contains no indictment against Aldus.

Aldus’s methods in the newly discovered parts of Books VIII, IX, and X The result of this exposure, Professor Merrill declares, should convince “any unprejudiced student” of the question that “Aldus stands clearly convicted of being an extremely unsafe textual critic of Pliny’s Letters.”⁷⁶ “This conclusion does not depend, as that of Keil necessarily did, on any native or acquired acuteness of critical perception. The wayfaring man, though a fool, need not err therein.”⁷⁷ I speak as a wayfarer, but nevertheless I must own that Professor Merrill’s path of argument causes me to stumble. I readily admit that Aldus, in editing a portion of text that no man had put into print before him, fell back on conjecture when his authority seemed not to make sense. But Merrill’s lists need revision. He has included with Aldus’s “willful deviations” from the true text of P certain readings that almost surely were misprints (218, 12; 220, 3), some that may well be (as 217, 28; 221, 12), one case in which Aldus has retained an error of P while I emends (221, 11), and several cases in which Aldus and I or i emend in different ways an error of P (222, 14; 226, 5; 272, 4—not 5). In one case he misquotes Aldus, when the latter really has the reading that both Merrill and Keil indicate as correct (276, 21); in another he fails to remark that Aldus’s erroneous reading is supported by M (219,17). However, even after discounting these and possibly other instances, a significant array of conjectures remains. Still, it is not fair to call the Parisinus Aldus’s only manuscript. We know that he had other material in the six volumes of manuscripts and collated editions sent him by Giocondo, as well as the latter’s copy of P. There could hardly have been in this number a source superior to the Parisinus, but Giocondo may have added here and there his own or others’ conjectures, which Aldus adopted unwisely, but at least not solely on his own authority; the most 65 apparent case of interpolation (224, 8) Keil thought might have been a conjecture of Giocondo’s. Further, if the general character of P is represented in Π, Book X, as well as the beginning of Book III, may have had variants by the second hand, sometimes taken by Aldus and neglected, wisely, by Budaeus’s transcriber.

The Morgan fragment the best criterion of Aldus With the discovery of the Morgan fragment, a new criterion of Aldus is offered. I believe that it is the surest starting-point from which to investigate Aldus’s relation to his ancient manuscript. I admit that for Book X, Avantius and the Bodleian volume in its added parts are better authorities for the Parisinus than is Aldus. I admit that Aldus resorted throughout the text of the Letters—in some cases unhappily—to the customary editorial privilege of emendation. But I nevertheless maintain that for the entire text he is a much better authority than the Bodleian volume as a whole, and that he should be given, not absolute confidence, but far more confidence than editors have thus far allowed him. Nor is the section of text preserved in the fragment of small significance for our purpose. Indeed, both for Aldus and in general, I think it even more valuable than a corresponding amount of Book X would be. We could wish that it were longer, but at least it includes a number of crucial readings and above all vouches for the existence of the indices some two hundred years before the date previously assigned for their compilation. It also supplies a final confirmation of the value of Class I; indeed, B and F, the manuscripts of this class, appear to have descended from the very manuscript of which Π was a part. We see still more clearly than before that BF can be used elsewhere in the Letters as a test of Aldus, and we also note that these manuscripts contain errors not in the Parisinus. This is a highly important factor for forming a true estimate of Aldus and one that we could not deduce from a fragment of Book X, which BF do not contain.

Conclusion I conclude, then, that the Morgan fragment is a piece of the Parisinus, and that we may compare with Aldus’s text the very words which he studied out, carefully collated, and treated with a decent respect. On the basis of the new information furnished us by the fragment, I shall endeavor, at some future time, to confirm my present judgement of Aldus by testing him in the entire text of Pliny’s Letters. Further, despite Merrill’s researches and his brilliant analysis, I am not convinced that the last word has been spoken on the nature of the transcript made for Budaeus and incorporated in the Bodleian volume. I will not, however, venture on this broad field until Professor Merrill, who has the first right to speak, is enabled to give to the world his long-expected edition. Meanwhile, if my view is right, we owe to the acquisition of the ancient fragment by the Pierpont Morgan Library a new confidence in the integrity of Aldus, a clearer understanding of the history of the Letters in the early Middle Ages, and a surer method of editing their text.

Nos. I-XII. New York, The Pierpont Morgan Library, MS. M. 462. A fragment of 12 pages of an uncial manuscript of the early sixth century. The fragment contains Pliny’s Letters, Book II, xx. 13—Book III, v. 4. For a detailed description, see above, pp. 3 ff. The entire fragment is here given, very slightly reduced. The exact size of the script is shown in Plate XX.

XIII-XIV. Florence, Laurentian Library MS. Ashburnham R 98, known as Codex Bellovacensis (B) or Riccardianus (R), written in Caroline minuscule of the ninth century. See above, p. 44. Our plates reproduce fols. 9 and 9^v (slightly reduced), containing the end of Book II and the beginning of Book III.

XV-XVI. Florence, Laurentian Library MS. San Marco 284, written in Caroline minuscule of the tenth century. See above, pp. 44 f. Our plates reproduce fols. 56^v and 57^r, containing the end of Book II and the beginning of Book III.

XVII-XVIII. Oxford, Bodleian Library, Auct. L 4. 3. See above, pp. 39 f. The lacuna in Book VIII (216, 27-227, 10 Keil) is indicated by a cross (+) on fol. 136^v (plate XVII^a). The missing text is supplied on added leaves by the hand shown on plate XVII^b (= fol. 144). The variants are in the hand of Budaeus. Plate XVIII contains fols. 32^v and 33, showing the end of Book II and the beginning of Book III.