A New Fragment of Plato's "Parmenides" on Parchment Willis, William H Greek, Roman and Byzantine Studies; Winter 1971; 12, 4; ProQuest pg. 539

A New Fragment of Plato's Parmenides 
on Parchment 
William H. Willis 

I N AUTUMN 1969 Perkins Library of Duke University acquired among 
a group of unrelated texts on papyrus a small palimpsest fragment 
from an ancient parchment codex containing Plato's Parmenides 

(PDuk. inv. G 5). Purchased from a dealer, neither its ancient nor its 
modern provenience could be ascertained, but we may confidently 
assume that it was found in Egypt. Measuring 7.6x 7.9 em., the parch­
ment was originally very thin and of high quality, but when acquired 
was extremely brittle and crinkled. Moreover the lower writing in 
Greek had been so thoroughly expunged in antiquity as to be almost 
invisible in ordinary light; but the occasional traces of large round 
majuscule letters could be observed from the impression of the pen 
in the surface of the parchment. When fluoresced by ultra-violet 
light, each face of the fragment readily yielded its text (see PLATE 6). 

On each side the fragment bears eleven lines of the Parmenides­
on the recto (flesh side) p.152B Stephanus, on the verso (hair side) the 
end of 152e and beginning of 152D. It brings the number of known 
ancient texts of Plato to 50,1 of which 44 are from papyrus rolls, 
6 from codices (3 of papyrus, 3 of parchment). The Duke fragment is 
the only ancient text of the Parmenides yet recovered, and is the 
earliest appearance of any of Plato's works in a codex or on parch-

1 So far as I have encountered them. R. A. Pack, The Greek and Latin Literary Texts from 
Greco-Roman Egypt2 (Ann Arbor 1965) [hereafter cited PACK2], recorded 44 published prior 
to April 1964-40 from rolls, 2 from III-cent. papyrus codices, 2 from later parchment (see 
n.2 infra). To this list may now be added (all from papyrus rolls unless otherwise noted): 
POxy. 2662 (Meno, I-cent. B.c./I-cent.); POxy. 2663 (Cratylus, later II-cent.); POxy. 2751 

(Republic, probably late II-cent., pOSSibly early III); PLeid. inv. 22 (Phaedo, I-cent.: in Mnemo­
syne 29 [1966] 231-40, 269, and cf Cd'E 42 [1967] 2Bf); and PAntin. 181 (Minos, early III-cent., 
papyrus codex). POxy. 24 (Pack2 1422) has been republished as PYaie 21; of PSI 1200 (Gorgias, 
Pack2 1414) a fourth fragment is published in Nuovi papiri letterari Fiorentini (Pisa 1971) 
no.9; and of the Rainer codex of Gorgias (PVindob. G 2600l=Pack2 1415) a new fragment 
(G 39880) is published in WS 78 (1965) 40-44, now redated II/III-cent. Even so, the new finds 
do not change the relative frequency of Plato papyri from Egypt (see GRBS 9 [1968] 211ff): 
Plato remains sixth after Homer, Demosthenes, Euripides, Hesiod and Callimachus, but 
increases his lead over Isocrates, in seventh place. 

539 



540 A NEW FRAGMENT OF PLATO'S PARMENIDES 

PDUK. INV. G 5 

PLATO, Parmenides 152B-D 

FLESH SIDE (recto) 

1 Tn 
2 NAI 
3 E C TIN l1 E n P E C 8 Y T€ P 0-
4 A P'O Y x 0 TAN KAT A T[o -] 
5 NYNXPONONHlrl 
6 rNOMENONTONME 
7 TA2YTOYHNTEKAI 
8 ECTAIOyrAPnOynO 
9 p E Y 6 MEN 6 N r E E K T[oY] 

10 nOTEEICTOEnEITA 
11 y n E p 8 H C ETA ITO N[Y N] 

HAIR SIDE (verso) 

1 MEN 0 N [T]O Y TEE n E I 
2 TAKAITOYNYN 
3 AAH8H· 
4 [E]I l1 ErE A N ArK H M H ~ 
5 nAPEI\8EINTONYN 
6 TIANTOrlrNOMENo-. 
7 ETIEIl1ANKATATOY 
8 T 0 A I E TI f C X E I A E I ToY 
9 rlrNEC9AIKAIECTI-

10 TOTETOYTo5TIAN~ 
11 [T]Y X H I r I r NOM E N 0 - . 

... 0;;-] 
7'w. 

Nat. 
"Ecn{v} 8~ 7TPECf3V" 7'E' pov 

1'" , tI , ['] ap ovx O7'av Ka7'a 7' OV 
A I 'I' 

VVV XPOVOV TJt yt-
I , 

YV0f-LEVOV 7'OV f-LE-
7'~V 7'OV ~V 7'E Ka~ 
" ,I 
EC7'at; OV yap 7TOV 7T0-

pEv6f-LEV6v yE €K 7'[ov] 
" ,,, 

7T07'E EtC 7'0 E7TEt7'a 

{mEPf3~CE7'at 7'0 V[ VV ]. 

3 EC7't B et om., EC7'W PDuk. 
5-6 YW6JLEVOV z. 7 ~v] vVv z. 
10 7'0] 7'OV n D R. 11 7'0] 

7'OV D. 

••• ytyv6-] 

f-LEVOV. [7' ]ov 7'E €7TEt-
, A A 

7'a Kat 7'OV VVV. 

• AA7J8fj. 
El 8' yE avaYKTJ f-L~ 

7TapEA8E'iv 7'0 vVV 
A' I 

7Tav 7'0 ytYV0f-LEVOJl. 
€7TEt8av Ka7'U 7'OV-

I _ t:I ... 

7'07'E 7'OV7'O 0 n av 

[7' ]VX7J' ytyv6f-LEVOV. 

-------------------- ------
7 rnE~ Mv z. 7-11 om. (91. 
9 €C7'W PDuk. et B. 10;n.L 
«Xv] 0 «Xv D R. 11 -roX7J'] WXOL 
E WI V fJ Heindorf. 



WILLIAM H. WILLIS 541 

ment.2 Its nearest mediaeval successors are the IX-century Paris 
MS grec 1807 (Burnet's A) and Codex Clarkeianus 39 (Burnet's B). 

The text of the Duke fragment is in virtually complete agreement 
with the Clarkeianus, on which modern editions are chiefly based. 
The only variant is a superfluous nu-movable in recto line 3, ECTIN 
l:1E, for a second superfluous nu-movable-a bar-nu at the end of 
verso line 9-is retained by B. Such nus before consonants are so 
common in papyri that we may wonder whether our Byzantine rule 
against them may have been too strictly drawn. Opposite are set 
forth in parallel columns a diplomatic transcription and an edited 
text; disagreements with the later MSS are recorded from the colla­
tion ofBrumbaugh.3 In all other readings the Duke fragment stands 
correctly with B against the later MSS; significantly at verso line 11 it 
has 7VXf}t against 7VXOL of the important XII-century Vienna 
codex (Burnet's W) and others. Offering no surprises, our text 
stands in welcome confirmation of the textual tradition. 

Except for one error, where in recto line 3 he omitted 7€ in 
nPECBYTE'PON, corrected above the line in a similar hand pre­
sumably by the SLOp(jW7~C, our scribe appears to have been meticulous. 
Iota-adscripts are written correctly and uniformly. Particularly 
interesting is his consistency in the use of accents and breathings. 
Acute accents occur twice in recto line 9 and once in verso line 4 to 
mark ensuing enclitics, and once each in recto 3 and verso 8 to forestall 
ambiguity in position or form of the verb. Similarly, the circumflex 
is used twice as an aid to the reader (recto 5, verso 8), in both cases to 
mark the subjunctive of €lpL A rough breathing (Turner's <form 1')4 

2 For an even smaller IV -century parchment fragment of Ps.-Plato, containing the end 
of Eryxias and the beginning of Demodocus (PackS 1429), see H. Hunger. "Pseudo-Platonic a 
in einer Ausgabe des 4. Jahrhunderts," WS (1961) 40-42: it is written in 'Biblical uncial' 
in the style of Codex Sinaiticus. Professor Hunger was the first to remark that the hand 
of the Duke Plato is not this, but "gehortjener Gruppe der fruhen 'romischen Unziale' an, 
die durch den Papyrus Hawara (Homer) und den Pap. Oxy. 20 reprasentiert wird, und 
konnte daher nach unseren bisherigen Kenntnissen ins 2. Jahrhundert gesetzt werden" 
(per !itt. 25 October 1969): judging from a photograph, he estimated the hand of the Duke 
parchment as "eleganter und asthetisch hoher stehend" than that of the Vienna (National­
bibliothek G 39846). The third parchment fragment is PAntin. 78, Theaetetus (packS 1394). 
assigned by its editor to the VI century. 

8 Robert S. Brumbaugh, Plato on the One (New Haven 1961) 305-06. To Professor Brum­
baugh I am indebted for discussing the text with me per litteras. 

'Eric G. Turner, Greek Manuscripts of the Ancient World (Oxford and Princeton 1971) 
14 [hereafter cited TURNER]: on the use of this and other signs employed by the Plato 
scribe see Turner pp.1D-14, and on the bar-nu p.17. All features noted are paralleled in 
papyri of the second and third centuries. 



542 A NEW FRAGMENT OF PLATO'S PARMENIDES 

appears once (verso 10), with accompanying acute accent, to dis­
tinguish two adjacent omicrons; and the uncertain mark over the 
ensuing alpha may be a smooth breathing. These signs are sparingly 
but consistently employed, therefore, in every case only where they 
are needed to prevent a reader's misunderstanding, according to the 
Alexandrian principle 7TPOC a,acroA~v rfjc a/-up,{36AOV Mgc:wc. The only 
instance of elision (recto 4 AP') is marked with an apostrophe; the 
trema appears over upsilon in YnEP (recto 11). Whether the scribe 
used punctuation on the recto remains uncertain, but on the more 
legible hairside we can be fairly sure of three single points: two, in 
middle position (rC:Ac:la cnYJ-L~), as full-stops closing speeches in lines 
3 and 11; and one, on the line ({J7TocrtYJ-L~), for a half-stop closing line 6. 
Angular spacefillers complete short lines at verso 4 and 10. At the 
end of long lines, nu is written as a supralinear bar, after omicron 
(verso 6 and 11, and presumably in lacuna at recto 3) and after ECTI 
(verso 9). Altogether, then, our text was carefully inscribed. 

A striking feature of format is ekthesis,5 projection into the left 
margin of the first letter to mark the beginning of each new speech 
in the dialogue. Each speech thus begins on a new line, even when 
its predecessor consisted of only one word, e.g. NAI (recto 2) and 
AAH8H (verso 3): there was no economy of parchment. This is not 
the format characteristic of the dramatic papyri nor of other papyri 
of Plato, which regularly use the paragraphus and double dots to 
mark a change of speaker. 

By calculating the length of the lacuna between recto and verso 
we may estimate the dimensions of the leaf from which our fragment 
derived and gain some idea of the original appearance of the codex. 
Between the carefully regulated right and left margins each line is 
6.8 cm. in length (not counting ekthesis), accommodating from 13 
to 17 letters, usually 14-16, at an average of 15 letters to the line. The 
letters, virtually circular or square, are approximately 4 mm. high, 
the interlinear space 2.5 mm.; vertical height of line and interlinear 
space is thus 0.65 cm. The text lost in the lacuna would therefore have 
occupied 18 lines which, when added to the 11 preserved lines of the 

5 Turner, op.cit. (supra n.3) 9. Ekthesis is a frequent feature of commentaries; an example 
showing projection by precisely one letter into the left margin is POxy. 853, a Thucydides 
commentary of the mid-second century, discussed by Colin H. Roberts, Greek Literary 
Hands 350 B.C.-A.D. 400 (Oxford 1956) no.17a [hereafter cited ROBERTS]' The same plate 
shows the use of angular space-fillers of the same type found in the Duke Plato. 



WILLIAM H. WILLIS 543 

recto, filled a column of 29 lines measuring 19 cm. in height by 6.8 cm. 
in width.6 Since such a column is too tall and narrow for a single­
column page, we may assume that the codex was written in double 
column, and that the Duke fragment derives from the outer column 
of its leaf. 7 

How generous were the outer and intercolumnar margins we can 
only guess. The surviving centimeter of margin at the left of the 
recto and at the right of the verso is presumably intercolumnar, which 
may have been wider still. Were this space no wider, the written 
area would have measured ca. 19 x 15 cm. Allowing modest margins 
of 2 cm. inner, top and outer and 3 at bottom would yield a perhaps 
minimal leaf size of 24 cm. high by 19 wide-a page very slightly 
higher and 4 cm. wider than that on which these words are printed. 
The margins may of course have been more gracious, but the relative 
proportion is likely to have remained the same, approximately 
5 : 4. This is just the proportion exemplified by the two earliest 
Greek parchment codices which have yet come to light, though both 
are somewhat smaller: the British Museum fragment of Demos­
thenes, De Falsa Legatione (19.5 x 16.5 cm.), also written in double 
column, as one would expect of prose; and the Berlin fragment of 
Euripides' Cretans (14.5 x 11.5 cm.), in single column.s Both are dated 
to the second century. Among papyrus codices are several of approxi­
mately similar size and proportion, ranging in date from the second 

S The text lost between recto and verso comprises 255 letters in Burnet's text (OCT); 
divided by the 15-letter average, it would require 17! lines. In the format of our text, the 
words immediately follOWing our recto, au ya.p (5 letters, one-third of a line), are spoken 
by the interlocutor; the rest of the line would have been vacant. 

7 Our fragment could not have come from the inner of two columns, for the 18-line 
lacuna would thus have been divided between two intervening outer columns of only 
9 lines each, shorter than the preserved 11 lines on each face. A three-column format (as 
in the Codex Vaticanus) would have yielded a written page 19 x 22 cm. and, with margins, 
a leaf wider than high (ca. 24 x 27). While such dimensions may have been possible, the 
great majority of ancient codices were taller than wide, even those that were virtually 
square. 

s Demosthenes: PLit.Lond. 127=Pack2 293, illustrated in New Pa/aeographica/ Society, 
Facsimiles of Ancient Manuscripts SBR.I pt.I, p1.2; its hand is small, practiced but somewhat 
irregular and not calligraphic; columns of 36 lines measure 14 by 5.5 cm.; top margin 
2 cm., bottom 2.7, intercolumnar 1. Euripides: BKT 5.2.73-79= Pack2 437, illustrated in 
W. Schubart, PGB tab.30a, and NPS SBR.2 pt.2, p1.28; though different from that of the 
Demosthenes, its hand too is small, regular but not calligraphic; top and bottom margins 
2 cm., left margin 1.5. 



544 A NEW FRAGMENT OF PLATO'S PARMENIDES 

to the fifth century;9 but so different are the methods and re­
quirements of manufacturing a codex from papyrus that compari­
son in size cannot be trusted to suggest the date of a parchment 
codex. 

At the scale of our fragment, the complete text of the Parmenides 
would have required 87 pages or 44leaves.1o To judge by our limited 
experience of other early codices, a codex of 44 leaves would seem too 
thin to stand alone, and we may guess that it contained another 
work (or more) as well-not necessarily Platonic, for many early 
codices had strangely assorted contents. Dercyllides' and Thrasyllus' 
arrangement of Plato's dialogues into tetralogies was known already 
by the first century, a sequence to become standard in the Middle 
Ages. If our codex had contained the entire Third Tetralogy, com­
prising Parmenides, Philebus, Symposium and Phaedrus, 231 leaves would 
have been required-a content too ambitious to expect before the 
fourth century. Of course, other groupings of the dialogues, as in 
trilOgies by Aristophanes of Byzantium, were known in the Alex­
andrian and early Roman centuries.ll 

The most remarkable feature of the Duke fragment is the style of 
the bookhand in which it is written, the large round majuscule of 
which the most celebrated example is the Hawara Homer (PLATE 7), 

9 In a paper entitled "Questions concerning the Typology of the Early Codex" offered 
to the XIII International Congress of Papyrology in Marburg 2 August 1971 and soon to 

appear in the Acta, Professor E. C. Turner classified papyrus codices in twelve groups 
according to size and proportion. Among these the Duke Plato most nearly approximates 
Turner's Croup 4, especially the Chester Beatty Gospels (PBeatty 1= PU) of the third century 
and the Vienna Homer (PRainer 3.4=Pack2 933) of the fifth, both single-column. But 
Turner deliberately and properly excluded parchment codices from his typological study, 
since they rarely conform to papyrus codex patterns. Meantime the publication of Pro­
fessor Turner's 1971 Rosenbach Lectures on the typology of the codex is eagerly awaited. 

10 Calculations are projected from Burnet's OCT edition, 13 lines of which would require 
40 lines in the parchment codex. Thus the 1623 OCT lines = 4994 codex lines; with space 
allowed for a colophon, these conform to In! columns, or 87 double-column pages. In 
the unlikely event that each page bore three columns, 58 pages or 29 leaves would have 
been required. 

11 Calculated from Burnet's OCT edition, Philebus would require 146 double-column 
pages in the format of the Duke Plato, Symposium 113 and Phaedrus 116, for a total of 462 
pages. The complete Platonic corpus as we have it would have filled seven such codices; 
but there is no evidence from antiquity for uniform multi-codex sets. For the various 
groupings of the dialogues, see Diog.Laert. 3.61f; A.-H. Chroust, "The Organization of 
the Corpus Platonicum in Antiquity," Hermes 93 (1965) 34ff; and R. Pfeiffer, History of 
Classical Scholarship (Oxford 1968) 196f. 



\VILLIS PLATE 6 

PDuk. in\'. G 5, recto: Parmenides 152[~ 

PDllk. in\'. G 5, versa: Parmenides 152c-D 

PARCHME:,{T FRAGMENT OF PL:' .. TO AT DUKE UNIVERSITY 

(Ultra-violet pilOtograplls by Hendrik ,'an Dii!: Jr, actllal si~e) 



PLA TE 7 WILLIS 

THE H<\WARA HOMER, BODL. Ms. GR. CLASS. J. l(P) [PACK2 616] 

(Excerpt, appr(1ximately actual siC!') 



WILLIS PLA TE 8 

Fragment of Hellanic!ls, Atlantis epOxy. 1084= PACK2 459) 

Fragment of Hesiod, Catalogue (POxy. 2354= PACK2 508) 

FRAG:\1E:'-ITS OF SECO:'-lD-CENTURY PAPYRUS ROLLS I:'-I HO:\IERIC MAJl'SCl'LE 

(ilpproximately act!lal siz:e) 



PLA TE 9 W ILUS 

THE A~mROSIA'" ILIAD, BIBL. AMBR. MEDIaL. COD. F 205 P. INF. 

(Excerpt from f21 r, approximately actual size) 



WILLIAM H. WILLIS 545 

which is assigned to the period ca. A.D. 150.12 Turner calls the style 
<formal round capitals', to avoid the misleading implications inherent 
in its prevalent designation <Roman uncial'.13 In its finest examples 
it is perhaps the stateliest and most elegant of the Greek literary 
hands developed in antiquity, found chiefly in spacious and de luxe 
papyrus rolls of the principal classics of Greek literature-Horner's 
Iliad and Odyssey, Hesiod's Works and Days and Catalogue, lyric poetry 
and tragedy, Cratinus and Eupolis, Hellanicus, Isocrates, Plato, 
Demosthenes, Theocritus and Polybius. Collecting more than twenty 
examples, Professor Cavallo recently traced the rise of the style in 
the late first century, its flowering under Hadrian and the Antonines 
and its decline and disappearance at the end of the second century.14 
While the sequence of development is disputed and the approximate 
dates of individual specimens are estimated variously within the 
period, all examples of the hand on papyrus are assigned by their 
editors within the span of little more than a century, from the late 
first to the end of the second. Although no literary roll bears a date, 
a document from the Fayum dated A.D. 88 (PLond. 141) is written in a 
hand closely related to the literary rolls.15 External evidence for its 
career may be seen in two Attic inscriptions, the first, bearing the 
date A.D. 61/2, by its rudeness suggesting that a new style had not yet 
been acclimated to epigraphical use, the second, dated A.D. 217/8, 

12 Bod!. MS. Gr. Class. a. 1(P)=Pack2 616 (Iliad 2); for the most recent discussion and 
illustration see Turner pp.1, 3,25,27, 38 and p!.13. 

13 Actually, Turner calls it the "first type" of <Formal round' (see Turner p.25), which he 
identifies by reference to the Hawara Homer but to which he assigns no specific name; his 
other two types of 'Formal round' are 'Biblical Majuscule' (Cavallo's term, vice Grenfell 
and Hunt's 'Biblical Uncial': G. Cavallo, Ricerche sulla maiuscola biblica [Florence 1967]) and 
'Coptic Uncial'. The term 'Roman uncial' was coined by Daniel Serruys, "Contribution 
a l'etude des 'canons' de l'onciale grecque," Melanges Chatelain (Paris 1910) 494f, and reluc­
tantly retained by Cavallo (see infra n.14). (Serruys attributed the term to F. G. Kenyon, 
who often used 'uncial' for <majuscule', and 'Roman hand' and 'Roman style' of all the 
first and second-century hands, but never the term 'Roman uncial' so far as I can deter­
mine.) Properly speaking, 'uncial' is a term with meaning specific only to Latin palaeog­
raphy; and 'Roman', if it suggests a period, is too broad, but if it suggests Rome itself as 
origin of the hand, is without foundation. Yet a specific term is needed. Until a better is 
devised, I suggest 'Homeric majuscule', since it is best known from the Hawara Homer 
and is found most frequently in Homer MSS, e.g. POxy. 20, PTebt. 265, PSI 8, not to mention 
the Ambrosian Iliad. At least it will serve as a counterpoise to 'Biblical majuscule', and with 
as much validity. 

H G. Cavallo, "Osservazioni paleografiche suI canone e la cronologia della cosiddetta 
'Onciale Romana'," AnnPisa SER.II 36 (1967) 209-20 with 12 plates. 

15 Roberts pU2a with description. 



546 A NEW FRAGMENT OF PLATO'S PARMENIDES 

by its perfunctory regularity and compression suggesting the end of 
the style.I6 These stones will testify that the style had currency out­
side of Egypt, and may have been practiced not only in Athens but 
also in other cultural centers of the Roman Empire. We cannot of 
course know where the manuscripts found in Egypt were written, 
but it has been suspected that particularly those on parchment may 
have been imported from elsewhere.17 

Characteristic of this style are the capital form of the alpha with 
horizontal cross-bar, the lunate forms of epsilon and sigma which 
with theta and omicron are virtually circular, the mu with deep rounded 
saddle, the unusual xi (found chiefly in this hand) with the lower 
members in the form of the arabic numeral 2 (2), the strictly bilinear 
iota, rho, tau and upsilon whose vertical hastae rest on the lower line, 
and the extremities of straight strokes finished with serifs. All these 
characteristics the Duke Plato shares with the examples on papyrus­
with the letter-forms of the Hawara Homer, but particularly with 
POxy. 1084, a fragment of Hellanicus' Atlantis dated' to the second 
century, and POxy. 2354, a fragment of Hesiod's Catalogue dated to 
the first half of the same century (see PLATE 8). In the Hawara Homer 
and the Hesiod the letters are of the same size (4 mm. high) as the 
plato (those of the Hellanicus being slightly smaller, 3.5 mm.); 
and while the average interlinear space of the Hawara Homer and the 
Hellanicus is slightly less (2 mm. vs. 2.5), that of the Hesiod is precisely 
the same as in the Plato. Moreover, the Hesiod exemplifies the same 
forms and sparing use of accents, rough breathing and apostrophe. 

Certain features of the plato hand, however, are rarely found in 
these models, and have been associated rather with hands considered 
either earlier or later than the second century. One of these is the 
divided kappa (1<); but this feature is found (occasionally to frequently) 
in PLond. 141 (A.D. 88), PFay. 110 (a letter of A.D. 94, Roberts pl.11b), 
POxy. 659 (Pindar, Parthenia, Turner no.21), and in the following 
second-century papyri: POxy. 841 (Pindar, Roberts pl.14, the second 
hand), 844 (Isocrates), 1083 + 2453 (Sophocles, Turner no.28), 1084 
(the Hellanicus), 1090 (Hesiod), and PRyl. 60 (Polybius). Scribal devices 

18 (1) IG II/IDa 1990= Joh. Kirchner/G. Klaffenbach, Imagines Inscriptionum Atticarums 
(Berlin 1948) no.127. (2) IG II/rna 3764=IIA no.145. 

17 Turner p.21: "Are all parchment codices found in Egypt to be regarded as written 
at centres outside Egypt? The assumption is commonly made that in Egypt. the home of 
the papyrus industry, papyrus not parchment would be used. But it is a guess only." 
See also Turner. Greek Papyri. an Introduction (Oxford/Princeton 1968) 50f. 



WILLIAM H. WILLIS 547 

to extend a short line or curtail a long one in order to maintain an 
even right margin are naturally not found in verse, where irregular 
line-length is usual, and three-fourths of our examples of this hand 
are in verse. Of the book-roll fragments of prose, only three preserve 
their right margins. These (POxy. 844, 2260 and PRyl. 60) share with 
the Duke Plato both the wedge-shaped line-filler and the tiny 
omicron under the overhang of a preceding or following letter; 
examples of the latter in other second-century hands may be seen 
in a wax tablet at the British Museum (Pack2 2713, see Turner noA), 
POxy. 2751 (Plato, Republic), PLit.Lond. 127 (the parchment Demos­
thenes) and PLit.Lond. 132 (Hypereides, Roberts pl.13b)-also in the 
first-century Ninus Romance at Berlin (PBerol. 6926, Roberts pl.lla). 
Another line-shortening device, the final nu written as a supralinear 
bar, is found also in POxy. 2751 (the Republic), POxy. 842 (Hellenica 
Oxyrhynchia, Roberts pl.I7b), POxy. 843 (Plato, Symposium), PBodmer 2 
(the Gospel of John) and Chester Beatty Codex I (Gospels, see n.9), all 
of the second century or early third. No doubt we should find the 
bar-nu frequently enough in the 'formal round capital' hand of the 
same period if we had other examples in prose of which the right 
margins were preserved. In any case, all these less common features 
of the Duke plato are substantially paralleled in second-century 
manuscripts. 

Of all the characteristics of the Plato parchment, one may be 
thought to point to a date later than the second century, since it is 
best known as a feature of the fourth-century 'Biblical majuscule' 
and its derivatives. That is the contrast between thick and thin strokes 
of the pen. Vertical strokes and those slanting from upper left to 
lower right (except in nu) are thicker, horizontal strokes and those 
drawn from lower left to upper right are thinner; and curved letters 
show subtle transitions from thin to thick at their upper left and 
lower right quadrants. Some observers have seen no such distinction 
in the strokes of the Hawara Homer or in some other examples, and 
even in cases where contrast is inescapable, have minimized its signi­
ficance.1s In fact, I find no specimens of the hand which do not reveal 

18 Serruys, for example, says (op.cit. [supra n.13] 495): "Ce type d'ecriture ne comporte 
ni pleins ni delies. Certes les conditions du ductus provoquaient parfois des pesees differen­
tes de la main, pesees qui influaient sur l'epaisseur des traits, mais l'opposition des traits 
minces et des traits pleins n'est point un des buts du calligraphe." Cavallo adds (op.cit. 
[supra n.14] 210-11): "11 tratteggio si presenta talvolta sottilissimo, piu spes so tenuemente 



548 A NEW FRAGMENT OF PLATO'S PARMENIDES 

some degree of distinction between thick and thin: even in the Ha­
wara Homer (PLATE 7), witness the thin cap of pi and tau, tongue of 
epsilon, cross-bar of eta and theta, left leg of alpha and lambda, rising 
arm of kappa, base of delta, in contrast to vertical and right-oblique 
strokes. In many other examples (e.g. POxy, 1084 and 2354 [PLATE 8]) 
the contrast is still more marked; and in PSI 8, 1212 and 1213, POxy. 
20,844 and 1090 there is obvious gradation between the quadrants of 
curved strokes-as in the Duke plato. In some specimens the 'chiaro­
scuro' is more pronounced, in others less, but it seems present in all. 
This is a feature which most distinguishes what we may call the 
'Homeric majuscule' from its predecessors and prototypes, the round 
majuscule of the first century B.C. and the first century.19 And this is a 
feature, become stylized and more mechanical, which is bequeathed 
to the 'Biblical majuscule' in the later third and fourth centuries. 

In the plato parchment the distinction between thick strokes and 
thin and the smooth transitions in curves are scarcely more prominent 
than, say, in PSI 8 and POxy. 1084, but they are more precise and more 
consistent. The thick strokes are no thicker, but the thin strokes are 
slenderer. This precision, I suggest, is not a function of a later stage 
of development, but of the interaction of the same cut of reed pen 
upon the smooth, hard surface of parchment, rather than upon the 
grainy, rough and more absorptive surface of papyrus. Writing on 
parchment is capable of greater regularity of style and more elegant 
finesse, and these are precisely the qualities by which the plato may 
seem distinguishable from examples of the same hand on papyrus. 
Except for its consistent precision, I find no stylistic features in the 
plato fragment which are not paralleled in other second-century 
exemplars. There seems to be no reason, therefore, not to assign it 
too to the second century. 

Since the Duke fragment is a palimpsest, from which the original 
ink was thoroughly cleansed to make way for a later text, we might 
hope that the later writing faintly visible on one surface could provide 
some clue to the history of the codex, or at least a terminus ante quem 
for the original text. On the recto, the smoother flesh side has preserved 

contrastato; a tal proposito va detto subito che un ben determinato chiaroscuro nello 
spessore dei tratti non rientra tra Ie regole del canone ... Anche in questo caso tuttavia 
resta escluso, come si e detto, un accentuato chiaroscuro." 

19 Of numerous examples may be cited Turner nos. 15,21, 37, 39,41 and 56 (which he 
denominates 'Informal round hands'), and Roberts plates Sa, 9b, 9c, lOa, lla and 12a. 



WILLIAM H. WILLIS 549 

no evidence of later ink; presumably it was never reused. On the 
verso there are traces of six lines of large rude letters written in brown 
ink, in coarse and irregular letter-forms that may be paralleled as 
early as the fourth century and as late as the sixth; with diffidence I 
should place them in the fifth century. The fragment of text appears 
to be written in Coptic, if €B07\. is correctly read in line 2 and a phai in 
line 4: another characteristic Coptic letter may appear in the doubtful 
hori beginning line 1, but here a sigma is possible. Since the left edge 
of our fragment must be quite close to the left edge of the original 
leaf (the outer margin of perhaps 1 to 2 cm. stood to the left of the 
text), we have the beginnings of the six lines. The writing is so large 
that the complete text would presumably have covered the entire 
leaf; in any case lines 1 and 3 run off the fragment at its right edge 
(the center margin of the original Greek text). From the ultra-violet 
photograph reproduced in PLATE 6, which is much clearer than the 
parchment itself, I have in great uncertainty extracted the following 
tentative transcription :20 

I ~AP~~~~YC€T[ 

2 €~O7\. €N~P • [ 

3 NOT€AJ.TT~NT[ 

4 AyAAq NAI [ 

5 €TTOOY CAp [ 

6 plf:l€ [ 

I take the text to be a fragment of a private letter in Sahidic. The 
presence of Sahidic suggests that the codex of plato may eventually 
have found its way to a monastic library in middle or upper Egypt. 

20 To Mrs Elinor M. Husselman and to my colleague Professor Orval Wintermute I am 
indebted for their study of the photograph, patient consultation and advice; neither is 
wholly persuaded of the transcription I present here, but both agree that the text is 
Coptic. We cannot be sure what dialect is employed, but there is no indication that it is 
not Sahidic. Broken texts in Coptic are notoriously difficult to interpret since the agglu­
tinating syllables have different meanings in different combinations. I have therefore 
spaced words only in lines 2, 4 and 5 where other combinations do not seem feasible: 
line 4 would thus mean "They made (or did) it for me," and €1TOOY in line 5 would 
mean "today." In line 1.z (so Wintermute) could be read c, and)\. could be read N. For the 
second € in line 3 Wintermute would read c. The unwritten spaces at the ends of lines 
4 and 5 and the spaces of line 6 remain unexplained. 



550 A NEW FRAGMENT OF PLATO'S PARMENIDES 

There, worn out or discarded when perhaps three centuries old, it 
yielded one of its leaves at least to be expunged and appropriated for 
private use. Our fragment from this leaf is, so far as we know, the 
sole surviving evidence of this once beautiful manuscript. 

The discovery of a specimen of the <Homeric majuscule' on parch­
ment cannot fail to invite comparison with the only other extant 
example of this hand on parchment, the famous Iliad in the Ambro­
sian Library in Milan. Of this once magnificent codex containing the 
entire Iliad in approximately 380 leaves illuminated with perhaps 
180 paintings, large excerpts from 52 leaves survive, bearing 58 of 
the miniatures cut from the codex probably in the XIII century.21 
Only so much of the text survives as happened to stand inscribed on 
the reverse of the excerpted miniatures. This text is written in the 
style I have called <Homeric majuscule' (see PLATE 9), although with 
certain embellishments of the hand. 

So long as the Ilias Ambrosiana remained unique, various scholars 
have assigned it to a variety of dates from the third to the sixth cen­
tury, variously interpreting the features of its hand or the style and 
repertory of its miniatures.22 It is said to be written on thick parch­
ment in a brownish ink, which first came into use late in the third 
century.23 Its equally round letters are larger than those of the 
plato (5 mm. high vs. 4), the interlinear space is somewhat wider 
(3 mm. vs. 2.5), and the individual letters are more widely spaced in 
a rather mechanical regularity. The letter-forms of the Ilias are uni­
formly those of the <Homeric majuscule', but with certain notable 
differences. One of these is the suspension of a prominent rounded 

21 For a handsome facsimile edition in color see Ilias Ambrosiana, Cod. F. 205 P. in! 
Bibliothecae Ambrosianae Mediolanensis (U. Graf, Bern 1953). A facsimile in monochrome 
was published by A. M. Ceriani and A. Raffi, Homeri Iliadis pictae fragmenta Ambrosiana, 
phototypice edita (Milan 1905). 

22 From palaeographical considerations Cardinal Mai and Ceriani judged it III century, 
Wilamowitz III/IV, Kenyon considerably earlier than V, E. M. Thompson (on different 
occasions) possibly III, IV or V, Bartoletti later than III, Gerstinger late V, Cavallo (op.cit. 
[supra n.13] 112) early VI. Similarly, on the basis of style and content of the miniatures, 
defenders of every date from III to VI century can be found among art historians. The 
views of many are summarized by R. Bianchi-Bandinelli, Hellenistic-Byzantine Miniatures 
of the Iliad (Ilias Ambrosiana) (OIten 1955) 45ff, though from the point of view of one who 
strongly advocates the early VI century on art-historical grounds. 

23 The use of brown iron-compound ink, which was to replace the earlier lampblack 
carbon ink in the fourth century and after, is first attested in a dated document by POxy. 
2269, in A.D. 269. The ink of the Plato cannot be compared, since it was expunged. 



WILLIAM H. WILLIS 551 

finial from the tongue of epsilon, which appears in no second-century 
example but is paralleled in POxy. 2334 (Aeschylus' Septem), assigned 
by Turner to the III/IV century;24 it is quite different, however, from 
the heavier angular finials found on epsilon in the fifth-century 
Codex Alexandrinus and the sixth-century Vienna Dioscorides (dated 
A.D. 512) and Vienna Genesis, with which it has sometimes been 
compared. The tongue of epsilon and cross-bar of theta are below 
center. Thin strokes are hardly more than hairlines, and the contrast 
between thick and thin strokes, while analogous in disposition to 
that of POxy. 1084,2260, PSI 8 and the Plato, has become stylized and 
stiffly mechanical. It is this severity, more than any other feature, 
which seems to separate the !lias from the Plato and the second­
century papyri. 

On palaeographical grounds, then, there is no feature of the hand 
of the Ilias Ambrosiana which compels us to assign it to a date later 
than the late third or early fourth century. But it is clearly later than 
the Duke Plato, which belongs rather with the second-century ex­
amples on papyrus and provides us for the first time with a prototype 
on parchment for the Ilias. Already the Plato illustrates the effect of 
applying to parchment a calligraphic formal hand, one characteristic 
feature of which was regularity of contrast between thick and thin 
strokes. This feature per se should no longer be adduced as evidence 
of a late date. 

If these comparisons are convincing, the Duke Plato joins the rare 
company of the Berlin leaves from Euripides' Cretans, the London 
leaf of Demosthenes, and possibly the fragment in Florence of 
Hesiod's Catalogue (PSI 1383) as sole survivors of the earliest Greek 
codices on parchment. Only the Plato is written in <Homeric majus­
cule', but there is no logical reason why this most splendid bookhand 
should not have appeared early on parchment. Meantime the Duke 
Plato becomes the first ancient fragment of the Parmenides to come to 
light, the earliest manuscript of Plato on parchment, and the earliest 

2C Turner no.26; Roberts, its original editor, assigned it to the second century, Cavallo 
(op.cit. [supra n.14] 50 n.3) to the fourth. All other finials and serifs of the Ilias can be 
paralleled in one or another second-century papyrus, including the much maligned delta, 
which most commonly has a long untipped horizontal base like that of the Hawara Homer, 
but rarely a slight thickening at one or both ends like that of second-century POxy. 2260 

(pace H. Gerstinger, Die Wiener Genesis [Vienna 1931] 54ff, who takes the delta as chief 
evidence for the late fifth century). 



552 A NEW FRAGMENT OF PLATO'S PARMENIDES 

Plato in codex form. In pristine state, it must have been a prime speci­
men of those most beautiful books which Lucian praised in Adversus 
I d t 2 " t T? \ \ A , 1\ \ "t' ,<;: 'A ' " \' n oc urn ,oca 0 n.Ct/\lUVOC HC Kal\l\OC TJ 0 aOLOLJLOC -r'nKOC cvv ETTLJLEI\HCf 

Tfj 7TCXCn ;'YPcapav.25 

DUKE UNIVERSITY 

September, 1971 

25 Earlier versions of several parts of this discussion were presented at the meeting of the 
Southern Section, Classical Association of the Middle West and South, on 7 November 
1970 in Miami Beach; to the Columbia University Seminar on Classical Civilization, 
18 March 1971; and to the XIII International Congress of Papyrology at Marburg on 
3 August 1971; the Marburg paper, to appear in the forthcoming Acta, is concerned less 
with the plato and more with the Was Ambrosiana. I gratefully acknowledge a grant of 
funds from the Duke University Council on Research in support of this study.