Bibliotheca scriptorum Graecorum et Romanorum

Teubneriana
Aeschinis orationes (Bla). Ed. corr. (Schindel). In Vorb.
Alciphronis epistulae (Schepers)
Ammianl Marcellini rerum gestarum libri qui
supersunt (Gardthausen)
Vol. I. II.
Ampelii ber memorialis (Amann). Ed. 2. In
Vorb.
Andocidls orationes (BlaFuhr)
Anonymi Byzantini vita Alexandri regis Macedonum (Trumpf)
Antlphontis orationes et fragmeata
(BlaThalheim)
Aphthonii progymnasmata (Rabe) Nicolai
progymnasmata (Feiten) Ioannis Sardiani commentarium in Aphthonii progymnasmata (Rabe) Romani Sophistae
peri aneimenu libellus (Camphausen). Ed
corr. (Haase). In Vorb.
Apulei Platonici Madaurensls opera quae supersunt
Vol. III. De philosophia libri (Thomas)
Cum. add.

Caesarls, C. Iulii, commentarii rerum gestarum (Klotz)

Vol. III. Commentarii belli Alexandrini
belli Africi belli Hispaniensis. Fragmenta
Catulli carmina (Bardon). Edd. corr.
Choricii Gazaei opera (FoersterRichtsteig)
Ciceronis, M. Tulli, scripta quae manserunt
omnia
Fase.
2. Rhetorici libri duo (Strbel)
5. Orator (Reis)
42 Acadcmicorum reliquiae cum Lucullo
(Piasberg)
43. De finibus bonorum et malorum libri V
(Schiche)
44 Tusculanae disputationes (Pohlenz)
45 De natura deorum (Piasberg)
46 De divinatione. De fato. Timaeus (Ax)
47. Cato maior. Laelius (Simbeck) De
gloria (Piasberg)
Scholia in Ciceronis orationes Bobiensia
(Hildebrandt)
Corpus agrimensorum Romanorum (Thulin).
Cum add.

Scholia in Aratum vetera (J. Martin)

Corpus hlppiatricorum Graecorum

Hoppe)
Vol I. II.

Archimedis opera mathematica (Heiberg et al.)

Dinaren! orationes (Bla)

Vol.
Vol.
Vol.
Vol.

I. Ed. corr. (Stamatis)

II. Ed. corr. (Stamatis)
III.
IV. ber einander berhrende Kreise
(DoldHermelinkSchramm)
Vol. V Index bibliographicus. Add. et
corr. (Stamatis). In Vorb.

(Oder

Diodori bibliotheca historica

(VogelFischerDindorf)
Vol. IV. VI (Index)

Aristotelis fragmenta (Rose)

politeia Athenaion (Oppermann)

Dionysii Halicarnasei quae exstant antiquitates Romanae (Jacoby)

Vol. IIV.
Supplementum. Indices
Vol. V. Opuscula critica et rhetorica
(UsenerRadermacher)
Vol. VI. Opuscula critica et rhetorica
(UsenerRadermacher)

Athenaei dipnosophistae (Kaibel)

Vol. IIII

Donati, Aeli, commentum Terenti (Wener)

Vol. IIII.

Augustini, S. Aurelii, confessiones (Skutella).

Ed. corr. (JrgensSchaub)

Donati, Tiberii Claudii, interpretationes Vergilianae (Georgi)

Vol. I.
Vol. II. Acc. Vitae Vergilianae (Brummer)

Aristaeneti epistulae (Mazal)

Caecilii Calactini fragmenta (Ofenloch)

Fortsetzung letzte Textseite u n d 3. Umschlagseite

Textual Criticism
and Editorial Technique
applicable to Greek and Latin texts
By Martin L. West
1973

B. G. Teubner Stuttgart

Dr. Martin L West

Born in London 1937. Educated at St Paul's School and Balliol.
Since 1963 Fellow of University College, Oxford. Editor of
Greek poetic texts, also author of a book on early Greek
philosophy and many articles on classical literature.

ISBN 3-519-07401-X (paperback)

ISBN 3-519-07402-8 (clothbound)
All rights reserved. This publication, or parts thereof, may not
be produced in any form, or by any means, electronic, mechanical,
photocopying, recording or otherwise, or transmitted without
written permission of the publisher
B. G. Teubner, Stuttgart 1973
Printed in Germany by Dr. Alexander Krebs, Hemsbach/Bergstr.
Cover design: W. Koch, Stuttgart

CONTENTS

Bibliographical Note, 5
PART I. TEXTUAL CRITICISM

1. Habent sua fata libelli, 7

Types of source, 9. The nature of manuscript transmission, 12.
Various causes of textual discrepancy, 15
2. Organizing the data, 29
Dealing with a closed recension, 31. Dealing with an open
recension, 37
3. Diagnosis, 47
The evaluation of variants, 48. Emendation, 53

PART II. EDITING A TEXT

1. Preparation, 61
Collecting the material, 62. Digestion, 68. The use of computers, 70
2. Presentation, 72
Prefatory material, 72. Choice of sigla, 74. The body of the
edition: general layout, 75. Text, 77. Between the text and the
apparatus, 82. The critical apparatus, 86. Some special types of
edition. Papyri, inscriptions, 94. Fragment collections, 95.
Scholia, 97. Indexes, 98. Printing, 101. Conclusion, 102
3

PART III. SPECIMEN PASSAGES

1.
2.
3.
4.
5.
6.

Hesiod, Theogony 176-200, 105

'Hippocrates', De morbo sacro 1, 29-44, 119
Aesop, fab. 157 Perry; Xenophon, Memorabilia 1, 3, 8-9, 128
Catullus 61, 189-228, 132
Ovid, Amores 3, 15, 141
Apuleius (?), De Platone 2, 20, 145

INDEX, 153

BIBLIOGRAPHICAL N O T E

This b o o k was written at the invitation of the publishers, w h o

wanted a replacement for Paul Maas's Textkritik (3rd ed., 1957)
and O . Stahlin's Editionstechnik (2nd ed., 1914). Stahlin's w o r k ,
the only detailed treatment of editorial m e t h o d , was excellent in
its day, but many of its r e c o m m e n d a t i o n s have been left behind
by fashion. Maas's w o r k will not date in the same way, for the
canons of textual criticism have long been established, and fashion
can only b r i n g aberrations or alternative f o r m u l a t i o n s ; but it is
too one-sided to be satisfactory as a general introduction. It
emphasizes the stemmatic aspect of textual analysis, and treats
contamination as a regrettable deviation a b o u t which n o t h i n g
can be d o n e , instead of as a normal state of affairs. I have tried
in Part I of the present manual to redress the balance, and given
some practical advice on dealing with contaminated traditions,
which I think is new. O t h e r w i s e there is little here that cannot
be found in other w o r k s on textual criticism, of which there are
plenty.
I could draw up a formidable list of such w o r k s if I t h o u g h t the
student o u g h t to read them. But textual criticism is not s o m e t h i n g
to be learned by reading as much as possible a b o u t it. O n c e the
basic principles have been a p p r e h e n d e d , what is needed is observation and practice, not research into the further ramifications of
theory. I therefore offer no formal bibliography, but content
myself with the mention of three b o o k s that will be referred to
several times in what follows.
1- Havet, Manuel de critique verbale appliquee aux textes latins
(Paris 1911).
5

G. Pasquali, Storia della tradizione e critica del testo (2nd ed.,

Firenze 1952).
H. Frankel, Einleitung zur kritischen Ausgabe der Argonautika
des Apollonios (Abh. d. Akad. d. Wiss. in Gottingen, Phil.hist. KL, Folge 3, Nr. 55, 1964).
Any of these may be read with considerable profit, especially
Pasquali's wise opus.

PART I

TEXTUAL CRITICISM

1. Habent sua fata libelli

Eduard Fraenkel in his introduction to Leo's Ausgewahlte kleine
Schriften recounts the following traumatic experience which he
had as a young student:
"I had by then read the greater part of Aristophanes, and I
began to rave about it to Leo, and to wax eloquent on the magic
of this poetry, the beauty of the choral odes, and so on and so
forth. Leo let me have my say, perhaps ten minutes in all, without
showing any sign of disapproval or impatience. When I was
finished, he asked: "In which edition do you read Aristophanes?"
I thought: has he not been listening? What has his question got
to do with what I have been telling him? After a moment's ruffled
hesitation I answered: "The Teubner". Leo: "Oh, you read
Aristophanes without a critical apparatus." He said it quite calmly,
without any sharpness, without a whiff of sarcasm, just sincerely
taken aback that it was possible for a tolerably intelligent young
man to do such a thing. I looked at the lawn nearby and had a
single, overwhelming sensation: . , . Later
it seemed to me that in that moment I had understood the meaning
of real scholarship."
Textual criticism is not the be-all and end-all of classical
scholarship, which is the study of a civilization. But it is an
indispensable part of it. By far the greater part of our knowledge
of that civilization comes to us from what the ancients wrote.
In almost all cases those writings have survived, if they have
7

survived at all, only in copies many stages removed from the

originals, copies of which not a single one is free from error.
Often the errors are so great that it is no longer possible to tell
what the author meant to say. It follows that anyone who wants
to make serious use of ancient texts must pay attention to the un
certainties of the transmission; even the beauty of the choral odes
that he admires so much may turn out to have an admixture of
editorial guesswork in it, and if he is not interested in the authenti
city and dependability of details, he may be a true lover of beauty,
but he is no serious student of antiquity. The dangers are obviously
more far-reaching if the text is being used as a source for historical
events, ancient life and manners, Greek or Latin linguistic usage,
or whatever it may be.
But the practice of textual criticism is more than a prophylactic
against deception. It brings benefits which go beyond its immediate
aims of ascertaining as exactly as possible what the authors wrote
and defining the areas of uncertainty. When scholars argue about
whether Aristophanes wrote or in such-and-such a passage,
the debate may seem trivial to the point of absurdity, and indeed
the sense may not be affected in the least. But by asking the
question "which in fact did the poet write?", scholars may be
led to inquire into the usage of the particles and the habits of
Aristophanes more closely than it would ever have occurred to
them to do otherwise. In the same way, by asking such questions
all the way through the text, they learn all kinds of things that
they did not know and never wondered about, sometimes things
that were not known to anybody. So our understanding of the
languages, metres, and styles of the Greeks and Romans has been
continually refined by the observations of clever critics. That in
turn helps us to form correct judgments about passages where the
sense is affected. This is to say nothing of the interest and value
that the study of such matters as the proclivities of scribes, and
the processes governing the spread of texts at different periods,
has in its own right.
Students have sometimes said to me that they recognize the neces8

sity of textual criticism, but they are content to leave it to the editor
of the text they are reading and to trust in his superior knowledge.
Unfortunately editors are not always people who can be trusted,
and critical apparatuses are provided so that readers are not
dependent upon them. Though the reader lacks the editor's long
acquaintance with the text and its problems, he may nevertheless
surpass him in his feeling for the language or in ordinary common
sense, and he should be prepared to consider the facts presented
in the apparatus and exercise his own judgment on them. He must
do so in places where the text is important to him for some further
purpose. This book, therefore, is not intended solely for editors,
but for anyone who reads Greek and Latin and desires some
guidance on how to approach textual questions. Textual criticism
cannot be reduced to a set of rules: each new problem calls for
new thought. But there are general principles which are useful
and not always self-evident, and these I shall try to explain.

Types of source
Most classical authors come to us in parchment or paper manuscripts which are seldom earlier than the ninth century and often
as late as the sixteenth. Some authors and works are preserved in
only one manuscript, in other cases the number may run into
hundreds. There are also a few cases in which early printed
editions serve as the only source, the manuscript(s) from which
they were made having since been lost. Sometimes different works
by the same author, or even different parts of the same work, are
contained in different manuscripts. Most classical manuscripts are
now in European libraries or museum collections, but some are
in monasteries (particularly in Greece) or private ownership, and
some are in such places as Istanbul or Jerusalem, or in American
libraries. Among the larger collections may be mentioned those
of the Vatican library, the Biblioteca Medicea Laurenziana in
Florence, the Ambrosiana in Milan, the Marciana in Venice,
9

the Osterreichische Nationalbibliothek in Vienna, the Bibliotheque

Nationale in Paris, and the British Museum in London.
For many Greek authors (and a few Latin) we possess also, or
only, remnants of ancient copies on papyrus or parchment, often
very small remnants but occasionally substantial. These continue
to be published year by year. They are very rarely older than
300 B. C. and never older than about 350. The largest number
date from the second and third centuries A. D., but they continue
into the sixth and seventh.
Apart from straightforward copies of complete works there
may exist excerpts in anthologies, epitomes, paraphrases, or
translations, e.g. of Greek works into Latin or Arabic. In some
cases these are the only sources extant, and even when they are
not, they may be of use as evidence for the text from which they
were made - which is not, of course, necessarily the true text, but
may well be an older text than that of surviving manuscripts.
Ancient or medieval commentaries and scholia hold out the same
promise. Their evidence about the text is of three kinds. First,
there are the actual quotations from the text, known as lemmata,
which serve as headings to sections of the commentary. Second,
there may be explicit statements of variant readings. Third, there
is the interpretation offered, which may presuppose a particular
version of the text. Unfortunately scholia are not usually documents with a definite date. They are added to as well as shortened
or altered in the course of time, and are liable to contain a mixture
of material of very different dates. At best they can inform us of
Alexandrian scholars' readings, at worst they testify only to the
obtuseness or perverse ingenuity of some medieval reader. It is
also worth noting that as they are usually transmitted together
with the text, the lemmata are liable to be adjusted to fit the
accompanying text, which may cause a discrepancy between
lemma and interpretation.
The evidence of ancient quotations, more surprisingly, is also
affected by this interaction with the direct manuscript tradition
of the author quoted. It might be thought that when one ancient
10

author quoted a passage from another, that passage would from

then on be preserved in a manuscript tradition quite independent
of the main one, so that agreement with the main one or with part
of it would take us back to the time of the quoting author. But
in practice it often happens that both traditions, that of the quoted
author and that of the quoting, show similar sets of variants. In
some cases these are such as could have arisen independently in
the two traditions, in others it is necessary to assume interaction *).
A well-known instance is the quotation of Virgil, Eel. 4,62, by
Quintilian 9,3,8. Virgil must have written qui non risere parentis nee
deus hunc mensa, dea nee dignata cubili est, and Quintilian's comment
on the change from plural to singular proves that he quoted it in
that form. But his manuscripts, like those of Virgil, give cui non
risere parentes, which must be an importation by a copyist familiar
with the already corrupted Virgil. A more dramatic example of
the possibilities of cross-contamination occurs in the Byzantine
historian Nicetas Choniata (p. 772 Bekker), who quotes some
lines of Solon that had earlier been quoted by Plutarch and in a
rather different form by Diogenes Laertius. One of the manuscripts
follows Plutarch's version, the other Diogenes'. Nicetas must have
taken the quotation from one of the two authors, but a copyist
who knew the other has seen fit to write it in - first no doubt as
a marginal variant, but subsequently put into the text. More will
have to be said presently on the limitations of quotations as evi
dence for the text.
Imitations and parodies are occasionally of use, especially in
the case of verse texts. For example, in Iliad 1,4-5, where the main
tradition gives ,
Athenaeus records that the pioneer of Alexandrian scholarship,
Zenodotus, read the more forceful instead of : it has
been observed that this is supported by the echo in Aeschylus,
Suppl. 800f. ' ' | . (But
*) See . R. Dodds, Plato: Gorgias, p. 64; W. S. Barrett, Euripides: Hippolytos,
Pp. 429 f.; my Hesiod: Theogony, p. 69 n. 1.
11

the imitation could not have been used to infer a reading if

its existence had not been recorded.) The text of an imitation, of
course, may in its turn have light thrown upon it by comparison
of the model. Thus in the case of Catullus 4 and the early parody
in Catalepton 10 of the Appendix Vergiliana, each poem is partly
restored from a corrupt manuscript tradition with the help of the
other.
The nature of manuscript transmission
Whenever a manuscript is copied, some mistakes will almost
certainly be made. But manuscript transmission is not simply a
mechanical process of cumulative error. The scribe may notice
errors in the exemplar before him and be able to correct them,
even without recourse to another copy; so it is quite possible
for his copy, the apograph, to be on balance more accurate than
the exemplar. On the other hand, the number of errors corrected
must always be less than the number made, and the overall trend
will necessarily be towards a less correct text. Besides, some of
the scribe's 'corrections' may themselves be mistaken, and this
kind of corruption is often more insidious than inadvertent
miscopying, being less easily detected afterwards.
The fact that errors occur in copying, and that the comparison
of different manuscripts brings variant readings to light, is no
modern discovery. It was well known in antiquity, as well as in
the Middle Ages, and the precaution was sometimes taken of
checking a newly-made copy not only against its immediate exem
plar but against another manuscript. When a variant was noticed,
it might be introduced into the new copy by correction, or it might
be noted in the margin or between the lines, preceded by some
such expression as (.), , , . ( = ),
al. (= alibi or aliter), vel. When a copy furnished with this kind
of primitive critical apparatus served in its turn as an exemplar to
another scribe, he might do any of four things. He might preserve
both the variant in the text (t) and the marginal variant (v) in their
12

places; he might retain t and omit v; he might adopt in place

of t, without mention of t; or he might put in the text and t in
the margin.
This confluence of readings from more than one exemplar is
known as c o n t a m i n a t i o n . When it is not present, the relationship
of copies to exemplars can be represented by diverging lines. For
example the diagram
A

expresses the fact that and C were copied from A, and D and
from B. But if the scribe of mingled readings from and C,
this calls for converging lines:

It is to be noted, however, that the line BE now represents a

different relationship from the line BD, namely a selection from
the readings that characterize B, instead of a more or less complete
reproduction of them.
If we were in a position to see the whole tradition of any ancient
author, that is to say if we had knowledge of every copy ever
made, and knew in each case which other copies the scribe used,
and if we had the patience and ingenuity (and a large enough sheet
of paper) to set out their relationships in the way just illustrated,
we could expect to see a complex system of lines ramifying from
the point representing the author's original. Most of the lines
would be diverging, and there might be considerable areas of the
diagram where they were all diverging, corresponding to periods
13

at which cross-checking was not customary, whether from scarcity

of copies or lack of awareness of the advantages. In other areas,
there would be much convergence. There would also be many lines
that came to a stop, corresponding to all the manuscripts from
which no copies were made. These would become increasingly
frequent as we approached the end of antiquity. In the case of
some authors the transmission would here peter out altogether;
for others it would be reduced to a single line, others again would
be a little luckier. Then, from the late eighth or ninth century
onward, the stream would begin to broaden out once more,
though it would never recover its former dimensions, and might
again run dry or be reduced to a trickle before the final salvation
of the Renaissance.
That is the sort of picture that must be held in mind. But of
all the manuscripts that existed, only a small fraction have survived, and often they all belong to the same small corner of the
whole diagram that we have imagined. If they happen to come
from an area where the lines are all diverging, we shall have what
is called a c l o s e d r e c e n s i o n , that is, it will be possible to construct a self-contained diagram (known as a s t e m m a ) which
represents the historical relationship of the manuscripts accurately
enough for useful conclusions to be drawn about the antiquity of
individual readings. The principles on which this is done are explained in the next chapter. If the extant manuscripts do not come
from such a straightforward area of the tradition, they may still
appear to, if not enough of them are extant to reveal the complexity of their true relationships; or it may be apparent that no
stemma can do justice to the situation, and we shall realize that
we are faced with an o p e n r e c e n s i o n 2 ) .
2

) The term and its opposite were invented by Pasquah, who also speaks of
'horizontal' as opposed to 'vertical' transmission when cross-contamination is
involved. Note that if only two manuscripts are preserved from a contaminated
area of the tradition, there will be nothing to show that it is contaminated: whatever errors they share can be attributed to a common exemplar. It needs a third
copy to show that things are more complicated.
14

But it is only if the number of extant manuscripts is rather small

that the recension is likely to be completely open or completely
closed. If there are twenty or so, it will probably turn out that
some of them are related simply and without contamination and
others not. If the older manuscripts can be fitted into a stemma,
the promiscuity of the younger ones should be easy to establish
and may be unimportant. If the relationship of the older manu
scripts resists analysis, it may still be possible to identify sub
groups whose structure can be stemmatized.
Various causes of textual discrepancy
Miscopying is far from being the only cause of textual variation,
and misreading is far from being the only cause of miscopying.
I conclude this chapter with a survey - which in the nature of
things cannot be exhaustive - of the variety of ways in which a
text may suffer change.
The first way is that the author himself may change it, after
copies have already gone into circulation. Aiistophanes revised
his Clouds after its production in 424/3, and both versions sur
vived into Hellenistic times (we have the revision) 3 ). The scholia
to Apollonius Rhodius' Argonautica are able to quote from an
'earlier edition' () of the poem, which differed here and
there from the current version 4 ). We have it on Ovid's own
authority that his Metamorphoses got into circulation before he
had fully revised it (Tristia 1,7,13 fT.), and our manuscripts of the
poem offer alternative versions of certain passages 5 ). The tradi
tions of several other ancient works have been shown or alleged
to betray the effects of issue at different stages of completion 6 ),
3

) See K. J. Dover's larger edition of the play, pp. lxxx-xcviii.

) See H. Frankel, Einleitung..., pp. 7-11.
5
) See, most recently, the brief but judicious remarks of A. S. Holhs, Ovid: Meta
morphoses Book VIII, pp. -xi. xxvii. 102-4. 117-8. 123-4.
6
) See Pasquali, Storia..., pp. 397-465; H. Emonds, Zweite Auflage lm Altertum
(Leipzig 1941); further bibliography in M. D. Reeve, Proceedings of the Cambridge
Philological Society 15, 1969, 75 n. 1.

15

sometimes on the ground of doublets standing side by side in the

text, sometimes on the ground of major divergences between
different branches of the tradition, if both versions convince the
connoisseur of their authenticity. In such a case it is not necessary
to assume that the two branches of the tradition have come
down quite independently of each other from the time of the
author 7 ).
There are others besides the author who may take it upon themselves to improve the composition. Greek tragedies suffered extensively from interpolations by actors (or at any rate for their
use), probably more in the fourth century B.C. than at any later
time 8 ); the plays of Plautus may have undergone something of
the sort on a smaller scale in the second century, but the evidence
is less clear 9 ). The embellishment of the Homeric poems by
rhapsodes is a similar phenomenon. It seems to show itself in
quotations by authors of the fourth century B.C., and in the
earlier papyri, which are characterized by additional lines of an
inorganic nature (often borrowed from other contexts) and some
other divergences from the vulgate 10 ).
Some kinds of text were always subject to alteration. Commentaries, lexica and other works of a grammatical nature were
rightly regarded as collections of material to be pruned, adapted
or added to, rather than as sacrosanct literary entities. When the
rewriting becomes more than superficial, or when rearrangement
is involved, one must speak of a new recension of the work, if not
of a new work altogether. The various Byzantine Etymologica,
7

) See W. Buhler, Philologus 109, 1965, 121-33, on Tertullian's Apologeticus.

) See D. L. Page, Actors* Interpolations in Greek Tragedy (Oxford 1934).
9
) See F. Leo, Plautinische Forschungen (1912 2 ; Darmstadt 1966), pp. 29ff. The
activity of later scribes may be involved, cf. C. C. Coulter, Retractatio in the
Ambrosian and Palatine Recensions of Plautus (Bryn Mawr, Pa. 1911). At some
date a second ending was composed for Terence's Andna; it does not appear in
all manuscripts, and did not appear in all those known to Donatus in the fourth
century.
10
) See Stephanie West, The Ptolemaic Papyri of Homer (Koln & Opladen 1967).

16

the treatises , and <, the proverb collec

tions, or any body of ancient scholia, will serve as examples 11 ).
The divergences between the two primary manuscripts of Longus'
Daphnis and Chloe show that copyists of this type of work too
felt themselves at liberty to change the wording as they went
along. In this case one cannot properly speak of different recensions;
but it is otherwise with the so-called Alexander Romance, which
exists in six different versions dated between 300 and 700 A.D.,
to say nothing of medieval and modern adaptations 12 ). The two
Lives of Aesop belong in the same category, and the various
collections of Aesopic fables. Again, at least parts of the Hippocratic corpus were subject to revision or rearrangement. Some of
the idiosyncrasies of an Ambrosian manuscript of the Oath have
now been proved ancient by P. Oxy. 2547; it does not follow
that the medieval alternatives are not also ancient. Another text
of a technical nature, Arams' Phaenomena, was rewritten in places
by Maximus Planudes (c. 1255-1305; see Bekker's apparatus at
lines 481-96, 502-6, 515-24); and as late as 1704 an editor of
Dionysius Periegetes' geographical poem thought it permissible,
in the interests of students, to omit and transpose certain passages
and to add new sections dealing with Muscovy, Tartary, America,
etc.
Changes of a less drastic but nevertheless dangerous kind may
arise when a passage is the subject of q u o t a t i o n . The main cause
is inaccurate memory, as it was the practice of most ancient writers
(apart from grammarians, who hunted for their examples in texts
as a rule) to quote short passages as they remembered them instead
of laboriously looking them up without the aid of numbered chapn

) See R. Reitzenstein, Geschichte der griechischen F^tymologika (Leipzig 1897);

M. L. West, T r y p h o n , De Tropis', Classical Quarterly n. s. 15, 1965, 230ff.
12
) See K. Mitsakis, Der byzantinische Alexanderroman nach dem Codex Vindob.
Theol. gr.-244 (Miinchen 1967), pp. 5ff\, and Ancient Macedonia (First Inter
national Symposium, Thessaloniki 1970), pp. 376 ff. The related work of Palladius
on the peoples of India and the Brahmans survives in two recensions; see the edition
by W. Berghoff (Meisenheim am Glan 1967).
17

ters or verses. (These are not the only alternatives: very often a
quotation is r e m e m b e r e d or copied from a n o t h e r a u t h o r w h o has
used it, or from an a n t h o l o g y . Hence a w h o l e series of quotations
can appear in t w o or three different writers, with the same distor
tions in each.) It was easy to make mistakes, to substitute or trans
pose w o r d s , or to conflate the passage with some other similar one.
Q u o t a t i o n s particularly tend to trail off inaccurately at the end,
as the q u o t e r ' s m e m o r y fails. A n o t h e r thing to beware of is that
he may deliberately adapt the construction or some other aspect
of the q u o t a t i o n to suit his o w n purposes. O n e type of adaptation
characteristic of a n t h o l o g y excerpts (which can be treated as a
sort of q u o t a t i o n ) consists in m a k i n g the passage m o r e selfcontained; for example has been substituted for at the
beginning of at least t w o excerpts from Solon in the T h e o g n i d e a
(153. 315), and a n u m b e r of excerpts that c a n n o t be checked from
fuller sources seem to have been altered at the end to fill out the
verse.
Christian zeal occasionally affected texts, as in the Vienna
manuscript of Ps.-Hippocrates , w h e r e the names of
G r e e k gods have in places been effaced, and replaced by
. T h e r e are cases of scribes bowdlerizing a text, that is, sup
pressing obscenity, t h o u g h it is surprisingly rare. H e r o d o t u s '
chapter on sacred prostitution, 1,199, is passed over by one
g r o u p of m a n u s c r i p t s ; one copyist of Martial toned d o w n the
vocabulary s o m e w h a t , substituting e.g. adulter for fututor and
turpes for cunnos in 1,90,6-7; and we k n o w of ancient critics w h o ,
shocked by Phoenix's admission that he seduced his father's
mistress at his m o t h e r ' s instigation, altered
to ' (II. 9,453).
Early G r e e k and early Latin texts u n d e r w e n t a natural process
of orthographical modernization in the course of time. T h e early
lonians w r o t e the contraction of and as , b u t the texts of
their w o r k s usually s h o w the later spelling ; and many a quoi
and quom has given way to cui and cum. But there was also a con
trary tendency, probably starting in late Hellenistic times, to try
18

to preserve or restore original dialect forms. T h e effect was to

introduce quantities of p s e u d o - I o n i c forms into the traditions of
H e r o d o t u s and the Hippocratica, and p s e u d o - D o r i c into that of
the bucolic poets. O t h e r w i s e it is difficult to point to examples of
systematic change p r o m p t e d by grammatical t h e o r y ; but Planudes
and his disciples regularly replaced , in the texts
they copied by , , and indulged one or t w o
other private preferences 1 3 ). M e n t i o n may be made here of the
p h e n o m e n o n that a certain manuscript or g r o u p of manuscripts
will sometimes consistently deviate in a fixed direction in a series
of places, not always for any apparent r e a s o n ; for instance, one
late copy of Apollonius R h o d i u s , Casanatensis 408, substitutes
for in 3,886. 1025 and 1098.
It is a general t r u t h that emendation by scholars and scribes is
much m o r e evident to us in the Middle Ages and Renaissance
than in antiquity, and at the same time that it constitutes a m o r e
serious p r o b l e m . T h e conjectures of ancient critics are some
times recorded in scholia and similar sources, but seldom appear
to have affected the textual tradition. T h e c o n t r i b u t i o n of any
individual must usually have been as evanescent as a pee into the
river. It was different in the M i d d l e Ages w h e n copies were few
and c o r r u p t i o n rife: e m e n d a t i o n was at once m o r e often called for
and more likely to c o l o u r the whole stream, or a conspicuous
branch of it. Scribes e m e n d e d what they could not read or were
unable to u n d e r s t a n d , and sometimes what was or seemed unmetrical. For instance, in the H o m e r i c H y m n 10,4-5,

,
the w o r d s underlined were not legible in the damaged copy from
which the oldest manuscript is derived, and ..
have been supplied instead, not altogether metrically.
In Lucretius 3,1,
13

) See my commentary on Hes. Th. 190. 480. 491.

19

 tenebris tantis tarn clarum extollere lumen,

the scribe of the copy from which all our manuscripts descend
left the initial /: to be done more ornamentally in red, but, as
often in such cases, the rubricator never got round to the job.
Of the two extant ninth-century copies, one reproduces the omis
sion, the other supplies O; some later copies supply A; another
again hits upon the correct Fi. In Juvenal 8,148,
ipse rotam adstringit sujflamine mulio consul,
we can see the corruption of mulio to multo leading to the metrical
'correction' multo sujflamine1*). Scribes' understanding of metre,
however, is seldom more than rudimentary, and often less. An
outstanding exception is Demetrius Triclinius (c. 1280-1340), who
had read Hephaestion's metrical treatise, and broke quite fresh
ground by venturing metrical emendations in Pindar and the
choruses of Greek drama on the basis of the strophic responsion.
Admittedly he made many mistakes even in the simplest metres 15 ).
We may now turn to the consideration of semi-conscious and
unconscious changes made by scribes in copying. It must be
emphasized that many of these are not visual but phonetic or
psychological in origin. When one is writing (whether one is
copying or not, but especially if one is), one tends to say the
words over to oneself. One may then find oneself writing down
a word that sounds the same as the one intended. Hence we find,
for example, as a variant for in Mimnermus 9,1 (the
Byzantine pronunciation of the two words being identical), and
similarly for ' in Theocritus 7,112 (evron). (The con
stant writing of e for ae in Latin texts, and for b (less often the
converse), is not really the same thing; most cases belong under
14
) Other examples from Latin poets are collected by . E. Housman, M. Manilii
Astronomicon liber primus, pp. lix-lxix.
15
) For a study of his emendations in Hesiod's Theogony see Class. Quart, n. s. 14,
1964, 181 f. (Add 94 for .) For those of a slightly
earlier, Planudean manuscript see ibid. 176f.

20

the heading of modernized spelling. But the implications for the

textual critic are similar.) Spoonerisms are not infrequent, i.e.
confusions like for , suscipit for suspicit. Thete is a
tendency to simplify consonant clusters, and to write e.g.
for , or astersi for abstersti.
The substitution of one word for another can be brought about by
m e n t a l a s s o c i a t i o n s of a non-phonetic nature; e. g. and
are often variants. A word that plays little part in a monk's
life may be mistaken for one that plays a greater part, e. g.
for . (Some examples from the tradition of Livy are
collected by R. M. Ogilvie, Greece & Rome n. s. 18, 1971, 32-34;
other Latin examples in Havet, Manuel.., pp. 263-264.) The
scribe may be reminded of a similar word or phrase that he has
copied earlier, it may be many pages earlier. Thus for , in Hes. Th. 454 the writer of gives , a
compound which has occured in 17 and 136. At Ovid Met. 12,103
inritamina cornu is corrupted to inritamenta malorum through remi
niscence of 1,140. Memories of particularly well-known authors
like Homer and Virgil were liable to intrude even without recent
copying being involved. The ends of verses or sentences suffer
most from this kind of error, because they regularly coincide with
the end of the piece of text that the scribe carries in his head while
his pen is in motion.
Because he carries a block of text in his head, at least a whole
phrase or half a line, he may unwittingly alter the order of the
words. One special type of transposition that occurs in Greek tragic
texts is the so-called v i t i u m B y z a n t i n u m , b y which a paroxytone
word is moved to the end of the iambic trimeter to make it sound
more like a Byzantine dodecasyllable. Parallel to this is a type
found for example in one family of Plutarch manuscripts, by which
the rhythm at the end of a sentence is adapted to Byzantine habits
(on which see P. Maas, Greek Metre, 1962, para. 23). Another type
aims to avoid hiatus in prose: a Ptolemaic papyrus of Xenophon's
Memorabilia (P. Heidelberg 206) reveals that the medieval tradi
tion has suffered from this kind of refinement. But a much more
21

general cause which operates in all kinds of text is the i n s t i n c t

t o s i m p l i f y . Bacchylides 15,47 w r o t e ,
/ ; but in the L o n d o n papyrus the last three w o r d s
stand unmetrically in the more straightforward order /
. O v i d Am. 1,14,1 w r o t e dicebam 'medicare tuos desiste capillos';
in some manuscripts this has become dicebam 'desiste tuos medicare
capillos\ Examples could be multiplied indefinitely 1 6 ).
This simplification of w o r d o r d e r is only o n e manifestation of
the tendency to banalize, to erode away the unusual form or ex
pression in favour of the everyday, becomes ,
becomes , dein becomes deinde, laudarit becomes laudauerit.
A s y n d e t o n is eliminated by the addition of a connecting particle.
T h i n g s that the a u t h o r left to be u n d e r s t o o d are made explicit.
Constructions that look incomplete are supplemented. It is often
difficult to say h o w far these processes are conscious. Perhaps
the copyist thinks he may as well make the text a bit easier to read.
Perhaps he genuinely thinks a mistake has been made. Perhaps
he writes d o w n w h a t he expects to see, w i t h o u t t h i n k i n g . He may
be misled by notes inserted in the margin or between the lines of
his exemplar to aid u n d e r s t a n d i n g , and take them as part of the
text or as corrective of it.
T h e b e s t - k n o w n species of this genus is the g l o s s , that is, a
w o r d or phrase that explains a w o r d or phrase in the text. Glosses
may intrude into the text, either in place of w h a t they were meant
to explain or in addition to it. F o r example, in H i p p o n a x fr. 72,7
' , the rare w o r d survives
only in one manuscript (in the c o r r u p t form ), while the
others give instead. In fr. 66 of the same poet, trans
mitted as , the
gloss has g o t in w i t h o u t displacing ().
A n o t h e r example, which will illustrate a different kind of gloss,
16

) See W. Headlam, Classical Review 16, 1902, 243ff.; W. Ronsch, Cur et quomodo hbrarii verborum collocationem in Ciceronis orationibus commutaverint,
Diss. Leipzig 1914; Haver, Manuel..., pp. 242-6.
22

is Petronius 36, quo facto uidemus infra {scilicet in altero ferculo)altilia

et sumina. A gloss that resembles the w o r d explained is particularly
liable to be mistaken for a correction; e.g. has ousted in the manuscripts of Aeschylus Ag. 282, and has
ousted , at Callimachus H y m n . 3,243. A particularly com
mon type of gloss consists of a p r o p e r name supplied where the
author used a circumlocution, as in the Panegyric of Messalla
(Corpus Tibull. 3,7) 56 Aetnaeae Neptunius incola rupis, where the
gloss Cyclops has i n t r u d e d into the line a b o v e 1 7 ) .
W h e n , h o w e v e r , a scribe writes a b o v e (Hes. T h . 161 L),
this cannot strictly be called a g l o s s ; n o r is it meant as a correction
or variant. It is advice to the reader o n h o w to interpret the
sequence of t h o u g h t . T h e writing of over vocatives in some
Latin manuscripts is not dissimilar 1 8 ). In these cases t o o , as with
glosses in the p r o p e r sense, i n c o r p o r a t i o n in the text will have a
banalizing effect. T h e r e are also kinds of marginal note which
are not glosses but can enter the text. T h e most i m p o r t a n t consists
of a q u o t a t i o n of some other passage that a reader has been re
minded of. So at Aeschylus Persae 253, the s o m e w h a t similar verse
Sophocles Ant. 277 is written in the margin of ; copies made
from have i n c o r p o r a t e d it in the text.
T h e r e are several ways in which an individual w o r d may be
miswritten w i t h o u t h a v i n g been misread. By far the c o m m o n e s t
way is partial a s s i m i l a t i o n to some o t h e r w o r d nearby. E n d i n g s
are particularly liable to be assimilated, b r i n g i n g confusion to the
syntax. T h e following examples occur in a h u n d r e d lines of Euri
pides' Heracles: 364 '
(for , assisted by phonetic equivalence); 372 (for ); 396
(for . . . ?); 398 ,'

') number of examples, and references to other discussions of the phenomenon,

are given by R. Merkelbach, Zeitschrift fur Papyrologie und Epigraphik 1, 1967,
100-2.
18
) See G. P. Goold, Phoenix 23, 1969, 198; Havet, M a n u e l . . . , p. 291.
23

 (for ); 412 (for );

441 (for ); 456 '
(for ). Medial assimilation is well illustrated
by the inscription on a sixth-century vase (Berlin F 1794): one
side has the correct : , the other has
: 11 OKA . No copying is involved here, which
proves that the error is entirely mental, not visual; and in a batch
of examination scripts which I had to mark in 1967 I noted no
less than 77 slips of the pen of the assimilative type, e.g. 'a critique
of the Roman of his time and of human nature in general* ('Rome'
assimilated to the coming 'human'); 'an excellent examplic of the
rhetoric of Gorgias'; 'bread, not oxen was the only food known
to Dicaeopolis which was put into an oxen' (for 'an oven').
Other standard types of psychological error are h a p l o g r a p h y ,
d i t t o g r a p h y , and simple o m i s s i o n . Haplography means
writing once what ought to be written twice, e.g. defendum instead
of defendendum; dittography is the opposite, reduplication of a syl
lable, word, or longer unit. My examination scripts produced
fourteen examples of dittography ('renonown' for 'renown', etc.,
but more often doubling of a short word like 'be' or O f ) , only
three of haplography. Omission too is especially liable to occur
with short words, insofar as it results from lack of coordination
between mind and hand. There is, however, another kind of
omission which may be of any length, and here we must make the
transition to error by misreading.
I refer to omissions committed by the scribe because he fails
to notice a portion of the text before him. Often the oversight
has a mechanical cause: similar words or phrases appear twice on
the same page, and the scribe, after copying as far as the first,
mistakes the second for the place he has reached, and so omits
what comes in between. This is the so-called s a u t du me me au
me me. For instance, in Seneca epist. 74,8 modo in banc partem,
modo in illam respicimus, one group of manuscripts omits the first
phrase, while in 113,17 repetition of the phrase ergo non sunt
animaiia has caused the omission of six sentences. In verse the
24

scribe's copying unit tends to be the line or half-line, so that

verbal similarities at line-end ( h o m o e o t e l e u t o n ) or line-be
ginning ( h o m o e a r c h o n ) 1 9 ) are particularly likely to mislead
him. In general it may be said of these mechanical omissions that
they affect short passages of a line or two much more frequently
than long ones, because the scribe usually remembers approxi
mately where he has got to on the page.
Simple m i s r e a d i n g of w o r d s is not uncommon. Letters
could be mistaken for other letters singly or in combination. The
particular errors liable to occur naturally varies according to the
different styles of writing current at different times. Familiarity
with these can only be acquired by examining manuscripts or
facsimiles, preferably with the guidance of an expert teacher 20 ).
But it will be in place here to list the standard letter-confusions.
Greek uncials (from 300 B.C.): A = = . = . G = (-) =
() = C. Il = e i . H = N = K = I C . AA = M. = . = V.
Greek minuscule (from the ninth century, at first concurrent
with capitals): = . = . = . = = . = . = .
= . = . = .
Latin capitals (down to the sixth century): = R. C = P.
C = G = O. D = . = F. = N. I = L = . = NI.
= . = Q. = .
,9

) The terms are applicable to any pair of words or phrases, not only to lines of
verse. For 'homoearchon' some use the equally correct formation 'homoearcton'.
'Homoeomeson', coined by Housman, has established itself as a useful addition
to the group.
20
) useful start can be made from E. Maunde Thompson, Introduction to Greek
and Latin Palaeography (Oxford 1912); C. H. Roberts, Greek Literary Hands
(Oxford 1955); E. G. Turner, Greek Manuscripts of the Ancient World (Oxford
1971); P. F. de' Cavalieri and J. Lietzmann, Specimina Codicum Graecorum Vaticanorum (Bonn 1910); F. Ehrle and P. Licbaert, Specimina Codicum Latinorum
(Berlin & Leipzig 1927, repr. 1967); R. Merkelbach and H. van Thiel, Griechisches
Leseheft (Gottingen 1965), Lateinisches Leseheft (Gottingen 1969), where further
bibliography.
25

Latin uncials (third to eighth century) : B = R . C = 6 = G = 0 .

CI = U. D = = U. F = = R. I = L = . = CO. =
. = Q.
Latin minuscule (from the eighth century): a = u. b = h.
c = e. cl = d. c = t. f = s . in = m = ui. = u.
It should be understood that this classification greatly over
simplifies, particularly with regard to Latin script: there were
several transitional forms between uncial and minuscule, and
some very different types of minuscule. In all the categories it is
the case that individual hands vary, some allowing more ambiguity
than others.
As long as a scribe finds his exemplar reasonably intelligible,
he does not read it letter by letter but takes in a whole word or
phrase at one glance. His misreadings are therefore not always
analysable in terms of individual letters: the aspect of the word as
a whole may deceive him if he happens to combine the strokes he
sees in the wrong way. This is especially easy in Latin minuscule,
where slightly sinuous uprights play an important part in the
formation of several letters. For example fitux (finit)\n Manilius
3,229 is misread as fixirr {sunt) ;mixc\nf{Mudus)
in 4,31astuueittf
{iuuenis).
Texts were written without word-division down to the end of
antiquity, and even later the division is sometimes incomplete or
inconspicuous. Many mistakes result from a copyist seeing part
of one word as part of another, or one word as two, etc.; e.g.
in Pindar Ol. 10,55 appears in some manuscripts
as ' . The rare word was hard to recognize. Similarly
in Propertius 2,32,5 deportant esseda Tibur, where has made
deportantes sed abitur. Obscure words and proper names fre
quently baffle the scribes. Catullus' annates Volusi (36,1) be
comes annuale suo lust - totally meaningless, but Latin words 2 1 ).
Copyists grasp at indications that they are writing the required
21

) Further examples in Havet, Manuel..., pp. 206-7.

26

language even if they cannot quite follow its meaning, and it is not
often that they abandon all pretence of articulacy. That happens
in the Medicean manuscript of Aeschylus' Supplices, where the
scribe had a single damaged uncial exemplar before him and was
often baffled by the difficult text. It happens most surely when a
Latin copyist who does not know Greek suddenly finds himself
faced with a Greek phrase or quotation: then he is reduced to
imitating the shapes of unknown letters, and gibberish soon
results, leaving the critic with beautiful examples of purely visual
corruptions unspoiled by other factors: AIAEOMAIIJACAHAHOAXPTCOIOMTICHNHC ( . Gramm. Lat. 6,505 Keil).
The use of a b b r e v i a t i o n s is a common source of error. In
antiquity abbreviation was little used except in documents and
texts of a sub-literary nature such as commentaries. The only
abbreviations common in literary texts are the numeral signs; a
stroke above the last vowel of the line, indicating final in Greek,
or m in Latin; and b', q\ for -bus, -que. Later, when Christian
scribes appear on the scene, we find the so-called n o m i n a s a c r a
abbreviated in the style (-)C for , DS for deus; declined, ace.
BN, DM, gen. (-)V, DI, dat. (%), DO. In time these abbreviations
came to be used in pagan texts too, and it is worth noting the ones
likely to be encountered there: BC = , KC = , YC =
, CTC = , ANOC = , OYNOC = ,
11 HP = (voc. \\E\\ ace. , gen. UPC, dat. , nom.
pi. 1IPKC etc.), HP = , COP = ^ =
(gen. I INC, dat. UNI, pi. 11 NAT A etc.)._DS = deus, S P | =
spiritus, DNS = dominus, SCS = sanctus, NR = noster (NRA,
NRI, etc.), VR = uester22). Derivatives can be formed from these,
e.g. = , = . It is this type of

22
) There are numerous variations. The classic study is L. Traube, Nomina Sacra
(Munchen 1907); more recent evidence in . H. R. E. Paap, Nomina Sacra in the
Greek Papyri of the first five centuries A. D. (Leyden 1959).

27

abbreviation that is responsible for such confusions as for

, for , domini for diu (Catullus 61,125).
The rise of minuscule script, with its cursive ancestry, brought
a much wider range of abbreviations into use, and no attempt
can be made here to give an account of them. Many of them involve
the replacement of letters (particularly in terminations) by short
hand symbols. Valuable works of reference exist on the subject 23 ),
but there is no substitute for experience with manuscripts if the
critic is to handle it with real skill. However, it is possible to
exaggerate its importance: abbreviations are not actually misread
as often as some ingenious emenders think.
A type of error that involves visual misinterpretation but not
actual misreading occurs when the copyist refers a marginal or
interlinear correction to the wrong place in the text. At Hesiod
Th. 239, for instance, L gives at the beginning of the line
instead of , but in the margin the scribe has written :
in one of the apographa this has been added to in 240, making
. It is much the same when an intrusive gloss displaces,
not the word being glossed, but another nearby (an example was
given on p. 23). It may also happen that a gloss or variant written
between the lines becomes conflated with the word below or
above, producing a mixture of the two which may be bizarre.
Thus at Hesiod Th. 355 the lost copy k must have had '- at
the end, mistakenly repeated from 353, with the correction
above it. The apographa and u preserved this arrangement, but
in the next generation the Mosquensis 462 (copied from K) pro
duced , and U (copied from u) . Again, if a phrase
or line is accidentally omitted and then restored in the margin,
the next copyist may insert it in the wrong place in the text: this
is the cause of many transpositions.
23
) See E. Maunde Thompson, Introduction to Greek and Latin Palaeography,
79ff.; W. M. Lindsay, Notae Latinae (Cambridge 1915); Dons Bains, Supple
ment to Notae Latinae (Cambridge 1936); . Cappelli, Dizionario di abbreviature
latine ed italiane (Milano 1929); A. Pelzer, Abreviations latincs medievales
(Louvain & Paris 1964) (supplements Cappelli).

28

Finally it must be noted that one corruption often leads to

another, some efford of interpretation on the part of the scribe
being usually involved. E.g. at Aristophanes Ach. 832 '
was wrongly divided as (R), and then became or
. Under this heading belongs the whole large class of scribal
emendations that are prompted by a corrupt reading and are them
selves mistaken. Multi-stage corruption of a purely graphic kind
is rare: where non-adjacent letters in a word or phrase have been
misread, as at Thucydides 6,74,2 became (-)PA(I)KAC, or as at Plato Gorg. 467b CXGTAIA became GXGTAIA,
it is often easier to assume that they were misread simultaneously.
In emendation, accordingly, one should not go too far in postu
lating multiple misreading.
The main causes of textual discrepancy have now been surveyed,
and it will be seen how various they are. They are not all equally
operative in any given text 24 ). On the other hand the critic must
keep in mind all those that are or may be operative in the text he
is dealing with, and not follow a one-sided approach. How he
may best set about his task is the subject of the next two chapters.

2. Organizing the data

The spade-work of collecting information about the readings of
the manuscripts and other sources is more likely to be done by
an editor than by anyone else, and I will deal with it in Part II.
For the moment I will assume that the critic has a body of such
information at his disposal.
His first job is to make an assessment of the quality of the dif
ferent sources, which is in part a question of their relationships
24

) 1 may mention here the copious collections of examples in Douglas Young's

articles 'Some Types of Error in Manuscripts of Aeschylus' Oresteia' (Greek,
Roman and Byzantine Studies 5, 1964, 85-99) and 'Some Types of Scribal Error
in Manuscripts of Pindar' (ibid. 6, 1965,247-73); W. Wyse, The Speeches of Isaeus
(Cambridge 1904), pp. xxxvi-xlvii; C. Austin, Menandn Aspis et Samia I (Berlin
1969), pp. 59-65; H. Frankel, Einleitung..., pp. 22-47.
29

to one another. Suppose there are, besides a medieval manuscript

tradition, some fragments of a papyrus a thousand years older
than any of the complete manuscripts, and a few ancient quotations. Can connexions be seen between the text of the papyrus
and that of the later copies? Are there any cases where the same
corruption is present in both? Are there cases where the papyrus
offers an inferior reading? Does it ever share an inferior reading
with part of the later tradition, and if so, does it side consistently
with a particular group of manuscripts? How were the quotations made, from direct knowledge or at second hand? From
memory or from a book? How scrupulous was the quoting author,
and how good was the text known to him? How reliable are the
manuscripts in which he is preserved, and do they give the quotation in the form in which he made it? Suppose again there is a
translation. When was it made? How accurate was it, and how
accurately has it been transmitted? Then there is the main manuscript tradition itself. Are any of the manuscripts directly descended from other extant copies? Is it possible to recognize
groups of closely related copies, or to construct a stemma? What
are the habits of the individual scribes?
The inquiry proceeds on two fronts, from external and from
internal evidence. It may be known from an external source who
made the translation, for instance. Manuscripts may contain dates
or signatures; if not, a palaeographer will be able to tell approximately at what periods they were written, and sometimes
where (in the case of Latin manuscripts, the regional varieties of
Latin script being firmly identified) or by whom. In dating paper
manuscripts watermarks can be a useful guide to the date of manufacture 1 ). Or there may be a record of a manuscript's original pro-

!
) See C. M. Briquet, Les filigranes. Dictionnaire historique des marques du papier
des leur apparition vers 1282 jusqu'en 1600. T. 14 (Leipzig 19232). - Opuscula,
(Hilversum 1955); The Briquet Album. A miscellany on watermarks, supplementing Dr. Briquet's Les filigranes, (Hilversum 1952). ( = Monumenta chartae
papyraceae histonam illustrantia, ed. by E. J. Labarre. Vol. 4 and 2.)

30

venance. Such data can be combined with what is known of the

general historical conditions that governed the transmission of
classical texts at different times 2 ), as well as with more particular
facts such as the movements of individual known scribes, or the
presence of the author in a medieval catalogue pertaining to an
identified library.
Upon this historical backcloth we project the more exact information derived from internal evidence, in particular the interrelationships of the copies as inferred from comparison of their
readings. How far they can be inferred, and how it is done, are
the questions that they will occupy us for the rest of this chapter.
1 am conscious of committing a hysteron proteron here, in
leaving for the next chapter the subject of the evaluation of
variants. Although this evaluation - which involves deciding not
only which variants are true and which false, but also which are
scribes' emendations - becomes easier after the character and
relationships of the sources have been defined, it is necessary, in
order that they can be defined, to carry out as much of it as possible beforehand. I think it will be better, nevertheless, if I postpone discussing its principles till later, and for the present simply
assume possession of the evaluative faculty.
Dealing with a closed recension
Whenever there are two or more manuscripts available, the attempt must be made to determine their historical relationship,
so that this can be taken into account, together with other considerations, in the evaluation of their variants. The attempt will
succeed approximately in proportion to the freedom of the tradition from contamination: 'the tradition', of course, means in this
context that area of the tradition which is represented by the extant
manuscripts.
2
) For a good brief survey see L. D. Reynolds & N. G. Wilson, Scribes and Scholars
(Oxford 1968).

31

In the absence of contamination, each copy will contain the

same errors that were in the exemplar from which it was made,
minus those that the scribe has seen and corrected, plus some
additional ones (unless, perhaps, the text is very short). This
axiom is the basis of stemmatic analysis. Suppose there are six
extant manuscripts, ABCDEF, related as follows ([a] [b] [c] denote
non-extant copies):
[a]

[c]
D

It will be possible to deduce their relationship from the pattern

of agreements and disagreements among them; only it is important
to realize that what is significant for this purpose is not agreement
in true readings inherited from more ancient tradition, but agree
ment in readings of secondary origin, viz. corruptions and emen
dations, provided that they are not such as might have been pro
duced by two scribes independently. The argument will run like
this:
There are some errors 3 ) common to all six manuscripts, there
fore they all descend from a copy in which all these readings were
present, unless one of the six is itself the source from which the
other five descend; but this is not the case, because each of them
has other errors which are not reproduced in the rest. The a r c h e
t y p e (defined as the lowest common ancestor of the known manu
scripts) is therefore a lost copy [a].
3
) In what follows 'errors' should be understood as a convenient abbreviation for
'readings of secondary origin' as just defined.

32

There are further errors common to BCDEF but avoided by A,

therefore they have their own < h y p a r c h e t y p e ' . As A has errors
of its own, this hyparchetype is not derived from A but inde
pendently from [a]. As BCDEF each have some errors which are
not shared by the whole group, none of them is identical with the
hyparchetype and we must assume another lost copy, [b].
BF share some further errors from which CDE are free. But there
are no errors in which are not also in F, while F has some errors
peculiar to itself. Therefore F is directly descended from B 4 ).
DE share errors from which BCF are free. Each also has pecu
liar errors, so neither is derived from the other, but both are
dependent upon another hyparchetype [c].
Of the three copies of [b], namely B, C, and [c] as reconstructed
from DE, no two agree in error where the third has the correct
reading and could not have got it by conjecture. No two of them,
therefore, are dependent upon one another or on a further hypar
chetype: all three come independently from [b].
Having established the stemma, we can use it to eliminate some
of the variants by showing that they originated in such-and-such
a manuscript and were not inherited from the earlier tradition.
Our aim will be to discover as far as possible what was in [a].
We shall ignore F, since we have found its exemplar; it can only
be of use to us as a source of emendations, or in a place where
B's text has become obscured or damaged since F was copied
from it. From the agreement of D with (in good readings as
well as bad) we can infer what was in [c]. If D and disagree, it
will be possible to decide which of them is faithful to [c] on the
basis of agreement with and/or C, provided that [c] reproduces
the reading of [b]; and even where it does not, it will often be

4
) In practice it is easy to mistake an apograph for a closer congener and vice versa.
The assumption of direct dependence is more certain if it is possible to point to
some physical feature of the exemplar which accounts for the reading of the apo
graph: a lost or torn page, words obscured by damp or written in a way that
invited misreading.

33

easy to decide which of the variants in D E must have arisen first

(in [c]) and which second. Next, from comparison of B, C and [c],
we shall usually be able to tell what was in [b], because it will be
seldom that more than one of them will diverge from [b] at the
same time, and then the agreement of any one of the three with
would give the reading of both [b] and [a].
We attain our goal, knowledge of the reading of [a], whenever
we find A in agreement with [b], i.e. with BCDE, BCD, BCE, BC,
CDE, CD, CE, BDE, BD or BE; or even with a single one of these
manuscripts where two copies of [b] have admitted innovations
independently. We can then treat this as the sole transmitted read
ing (to be judged on its merits), and disregard the other variants.
But what if A and [b] disagree? In some cases it may be clear
which reading was the source of the other. Otherwise the question
must be left open 5 ).
Now that the principle has been explained, it is necessary to
point out some possible complications that were deliberately
excluded from the hypothetical situation given above. Suppose
that of the nine manuscripts in the system only three were extant
instead of six, namely A, F and D. Comparison of their readings
would give us the following stemma:

[b]

A
F

) Sometimes the condition of an archetype allows deductions about earlier copies.

The Lucretius archetype, for instance, contained some errors caused by misreading
of Rustic capitals, others caused by misreading of insular minuscule, so that two
'proarchetypes' are conjured up.
34

We would judge FD to be two copies of [b], and so removed two

steps from [a]. We would not be able to tell that there were in
fact intermediate copies between [b] and F and between [b] and
D, unless we found corruptions such as could only have occurred
in two stages. Any stemma that we construct for the manuscripts
of a classical author is liable in the same way to be an oversimpli
fication of the historical reality. However, it will not be a serious
falsification, provided that our reasoning has been careful, and
provided that contamination is not present in the system. We shall
get the readings of [a] in nearly all cases from AFD as surely as
we would from ABCDE. Even if only AF are extant, their agree
ment will still give us [a], though we shall be worse off in that
they will disagree more often than A and [b] did, and so leave us
more often in doubt.
If contamination is present we may be seriously misled. Suppose
that the scribe of F, besides copying B, kept an eye on A and
borrowed some readings from there, and suppose then that A
was lost. The true stemma would now be:

[a]
[A]

[b]

[c]

C
D

We would observe that F sometimes avoided errors common to

the rest, and that sometimes sided with C and [c], sometimes
with F. We would construct this stemma:
35

We would discard as a contaminated manuscript offering

nothing that was not to be found in the other sources, and we
would treat the peculiar readings of F as being as likely as those of
[b] to be those of the archetype. Insofar as they were drawn from
A, this would be correct, but insofar as they were errors made
by B, or in copying from B, it would be false. If a codicologist
told us that was in fact written earlier than F, we would merely
postulate an intermediary between [a] and F, and adjust our
dotted lines.
Such misapprehensions no doubt occur, and there is no infal
lible way of avoiding them. The best one can do is opt for the
hypothesis that fits the facts most straightforwardly, taking great
care that it does fit them. If contamination is present in more than
a slight degree, it will be found that no stemmatic hypothesis is
satisfactory. Before we consider what to do in those circum
stances, let us notice one other kind of complication that may
affect a closed recension.
Suppose that the archetype [a] (in our original, uncontaminated
system) contained marginal variants or corrections, so that the
scribes of and fb] were faced with the various choices listed on
pp. 12f. We might find as a consequence an inferior variant shared
by with one of the three copies of [b], something that could
not happen, except by coincidence, with the simple form of lineal
descent that we began by envisaging. If we found that it had
happened, there would be two explanations available: (i) that
36

there were variants in [a], reproduced in [b], and the scribe of C

(let us say) made the same choice as the scribe of ; (ii) that the
link between and C is the result of cross-contamination, C
having consulted A or a copy closely related to , or vice versa.
There is a good chance that if (i) is the case, one or two of the
extant manuscripts will still give both variants. At Catullus 12,4,
for instance, the primary manuscripts, being related as shown,
give:
[V]

falsum
al. salsum

[XI
G
salsum
al. falsum

R
falsum
al. salsum

V must have had the variants; the agreement of OR does not

show that it had falsum in the text. Whenever several manuscripts
sporadically or simultaneously record each other's readings as
variants, this is a probable sign of variants in their exemplars.
When on the other hand a single manuscript offers a series of
marginal variants which correspond to those known from another
manuscript or family, it is a sign of contamination.

Dealing with an open recension

Besides OGR there are many later, derivative manuscripts of
Catullus whose affinities cannot be reduced to a stemma. But
because the three primary manuscripts permit us to reconstruct
the archetype, it is possible to concentrate on them and treat the
tradition as a closed one. The situation is similar with, for example,
Lucretius and Theognis. We speak of an open recension when
the older manuscripts, or more strictly all those manuscripts in
37

which worthwhile variants (other than emendations) appear for

the first time, are not related perspicuously and do not allow us
to construct an archetype.
Different kinds of situation may be involved here. Firstly, the
contamination may not be as total as it seems at first sight: some of
the manuscripts do descend from an archetype directly enough for
it to be reconstructed, and it is only the eclecticism of the others
that confuses the picture. In its simplest form this situation may
be represented thus:

Secondly, there may be an archetype from which all the extant

copies are indirectly descended, but with cross-contamination
making intermittent links between all the branches of the tradi
tion, so that their agreements are never reliable indications of
what was in the archetype. Thirdly, it may be that there was no
archetype at all, apart from the author's original, because two or
more unrelated ancient copies survived into the Middle Ages to
become independent fountain-heads.
How is the critic to discover which of these situations he is
facing? There is no infallible way, but certain criteria are ap
plicable. Suppose the first case is true, the one illustrated by the
stemma above. It is easy to see that whereas AC will sometimes
agree in error against B, and BC against A, AB will never agree
in error against C 6 ). Here lies a criterion. If we tabulate the
combinations in which the manuscripts err, and fill in the number
of agreements in significant error not shared by the whole tradi
tion, thus 7 ):

) I must refer again to p. 32 n. 3.

) The table cannot show agreements of three or more manuscripts, but that is
unnecessary for the present purpose. There are quicker ways of doing the job.

38

C
D

C
14
12

D
16
6
27

7
7
25
22

F
19
3
8
7
9

G
8
9
11
10
8
20

it will soon be seen whether there are any manuscripts that never
err together where another part of the tradition preserves the
truth. (Any very low figures will deserve closer investigation.) It
should be possible to put any such manuscripts in a serviceable
stemmatic relationship. For instance, if never errs with or P,
though and often err with each other, we can extract the
stemma

and discard all the other manuscripts, which have no better read
ings to offer (ex hypothesi) and merely confuse matters.
I spoke of a 'serviceable' stemmatic relationship, meaning 'not
necessarily historically exact'. Suppose we have six manuscripts
BCDEFM, actually related as in the stemma overleaf.
When their readings are compared it is obvious that BCEF form
a close-knit group; further inspection reveals its structure, and
we can now quote b instead of the individual copies. What is its

Instead of entering numbers, one may simply tick the spaces as the various agree
ments are established (more care must then be taken to see that they are significant).
Or, if it is once established that C alone preserves such-and-such a true reading, it
only remains to show that each of the other manuscripts is somewhere in error
with C. If each of three sources is somewhere the only one to preserve the truth,
the question is answered.
39

relationship to D and M? We observe that DM sometimes agree

in error against b, and b at other times against D ; but never
D b against M, for cannot preserve the truth by itself. If it got
it from --0 it would also be in D, and if it got it from it would
also be in b (which has no choice when and agree). We will
infer a stemma of the form

Now this is clearly rather more than an oversimplification of the

true state of affairs. But it expresses correctly the basic truth that
readings from which survive in extant manuscripts must have
followed one of two routes, and, barring two independent cor
ruptions, must appear in D or b.
40

By the use of the tabulation method, then, we can determine

whether there are lines of tradition that have remained more or
less independent of each other throughout. It does not tell us,
however, whether there was an archetype. Take the stemma given
above for BHP. It posits an archetype, and for practical purposes
the assumption will be harmless. But it is possible that derives
from one ancient copy and HP from another that was not cognate
with the first. The common assumption that the transmission of
a Greek or Latin author normally depends on a single minuscule
manuscript, the first and last to be transcribed from an ancient
copy or copies (with variants recorded in some cases), is of rather
limited validity, as Pasquali has demonstrated 8 ). What criteria can
be applied here?
It might be thought - and Pasquali himself thinks - that if all
the manuscripts agree in error that implies misreading of minu
scule script, there must have been a minuscule archetype. Not
necessarily. If there has been contamination, there is no certainty
that those errors were all to be found in the same prototype, or,
if they were, that it was an archetype in the sense that all its errors
were inherited by the later tradition. But the quantity and the
quality of the errors come into consideration here. The greater
their number, the less chance there is that they were diffused
'horizontally'; and again, the more obviously false they are, the
less credible it is that they were chosen by scribes to whom alter
native readings were available.
That there was no single archetype may be inferred in the fol
lowing ways, (i) From the presence in the medieval tradition of
many pairs of variants known to be ancient. reading can be
taken as ancient either if it is attested by an ancient source or if
it is true. If the medieval tradition in a series of places preserves
true readings which could not have been restored by conjecture,
side by side with false variants which already occur in papyri or
) S t o n a . . . , chapter VI, particularly pp. 210-3. 259-61. 273-8. 295-8. 303-4.
375-8. 386-9. See also Barrett, Euripides: Hippolytos, p. 58.
41

ancient quotations 9 ), and if this happens more than can reasonably

be accounted for by the recording of variants in the margins of an
archetype 10 ) or in a body of scholia, then it is necessary to assume
more than one line of transmission from antiquity, (ii) From the
presence of divergences so substantial, or at so early a date within
the Middle Ages, that one cannot believe them to have arisen in
the short time available or under the conditions that prevailed
after the end of antiquity.
In the absence of such indications it will often be an open
question whether there was a single archetype or not. As I have
intimated, it is not actually a question of much practical importance. What is important is to recognize that medieval variants
are often ancient variants, and that ancient attestation of one of
the manuscript alternatives does not necessarily mean that it is
the true alternative.
When the critic has established that no stemma can be constructed, how is he to proceed? He must, of course, see what groupings
are apparent among the manuscripts, and whether the individual
groups can be analysed stemmatically, as was the case with BCEF
in the example on p. 40; even if they cannot, he can treat them as
units in his further cogitations, provided that they have a sharply
defined identity. Thus he reduces his problem to its basic terms.
In distinguishing fundamental affinities from the superficial ones
produced by contamination, in other words, in distinguishing
vertical from horizontal elements in the transmission, he may
sometimes be helped by the principle that the most significant
agreements between manuscripts are those involving omissions
and transpositions (insofar as they are not due to some evident
mechanical factor such as homoeoteleuton), since these are not
easily transmitted horizontally - though it is not impossible.
9

) Provided that the quotations were unknown to the scribes. See also pp. 10-11
on the influence of direct traditions an the text of quotations.
10
) Barrett points out that such an archetype would have to contain far more
variants than the number preserved among its surviving descendants.
42

The aim now is to determine which of the manuscripts or manuscript families are most independent of each other, for these must
go back most directly to the earliest phases of the tradition that
we can reach, and they must be the most fruitful sources of ancient
readings. The critic will take note of the general appearance of
the various witnesses first - that this one is a humanist's copy,
liable to contain much emendation; that here is a major family
that preserves its identity from the twelfth century into the
fifteenth, and so on. His attention will naturally be drawn by
any manuscripts that stand out from the rest by reason of their
age, or are notable for agreements with an ancient source. He may
well consider whether these apparently noble documents are all
that he needs, or whether there are others that preserve further
ancient readings.
This approach perhaps savours of trial and error, but it is not
difficult to develop it into a generally applicable sorting procedure.
The steps are these:
1. Whenever the manuscripts are at variance, make a note of the
reading or readings that seem to be ancient (true, and not found
by conjecture, or else attested by an ancient source unavailable
to the scribes) and the manuscripts in which it or they appear.
2. Any manuscript that is the sole carrier of such readings is obviously indispensable. Adopt it.
3. Remove from the list all the readings for which the manuscripts
just adopted may serve as sources (not just the readings that
appear only in them).
4. See which manuscript contains the largest number of the remaining readings. Adopt it. Remove from the list the readings
it contains.
5. Repeat (4) until every ancient reading is accounted for.
This is the most efficient way of reducing to a minimum the
number of manuscripts that have to be quoted as witnesses to the
tradition. The remainder can be eliminated. The elimination is primarily for practical convenience: it is not quite like the elimination
43

of apographa which demonstrably cannot tell us anything new.

Nevertheless, there must be a correlation between lack of individual good readings and lack of independent sources. Even though
the sorting method cannot elucidate the complex affinities of the
manuscripts, it produces results that stand in some relation to
them. This may perhaps be made clear with the help of an imaginary example. Suppose we have fourteen manuscripts ABCDEF
G H I K L M N O , actually related as follows (the stemma is an extended form of the one on p. 40):

When their readings are compared it is soon discovered that N O

are copied from D and can be dispensed with; that BCEF are so
related that their hyparchetype b can be used instead; and that
G H I K L form a group which can be treated as a unit, even if its
structure resists analysis. The fourteen witnesses are thus reduced
44

to five, ADM b g. Further stemmatic construction turns out to

be impossible. None of the five is the direct source of another - not
consistently, at any rate - and each of them shares errors with
one, two or three of the others in bewilderingly various combina
tions. We apply the sorting procedure. We find that only has
good readings peculiar to it; b has the largest number of the good
readings that are not in A; and D contains all the remainder,
whereas there are some that are not in and others that are not
in g. We decide therefore to base our recension on A b D.
What can we guess about the sources lying behind our manu
scripts? A's unmatched ability to produce individual good read
ings (despite its late date) implies access to a source independent of
those that feed the others n ) . The others perhaps draw on the same
range of sources, but b and D have profited most from them.
and g have chosen less well, or found them in a less pure state;
each sometimes has a good reading that one or two of the trio
A b D have missed, but never one that all three have missed.
These conclusions approximate to the truth. A has individual
good readings because of its access to . When it is in the wrong,
the right reading is more often found in b than anywhere else,
because A followed more often than b, and followed -0
more often than the purer line -. When A and b are both in the
wrong, the right reading is preserved in D if at all, because, if a
corruption occurs (or is present from the start) on the line
y-$-e-(M-)b, and A fails to pick up the true reading from , then
it can be preserved only by way of and 0, and D is the only
manuscript consistently faithful to 0. cannot preserve the truth
against both D and b, for the reason given on p. 40, and a similar
argument applies to g.

n
) If there had been a second copy that also had good readings peculiar to itself,
that would not necessarily imply a second such source. It and might be selecting
independently from the same source. But in this case we should expect them at
least as often to make the same choice and agree in good readings unknown to the
rest of the tradition.

45

Now this manufactured example will illustrate a general truth.

Whenever the manuscripts give divergent readings, one of which
is primary and the other secondary, the secondary reading has
originated at a given point in the tradition, and has come down by
certain routes to all the copies in which it appears; the primary
reading has come down by a route or routes which bypass that
point and have no station in common with those other routes.
The extant tradition will depend ultimately upon a very limited
number of early medieval copies, and there will be a very limited
number of possible routes from them to the extant manuscripts
that do not touch at any point. To escape from the oldest corrup
tions we can escape from, we depend on those routes. The extant
readings will not have consistently come down by those routes to
particular manuscripts, in a contaminated tradition; but the sort
ing procedure that I have recommended must have the effect
of isolating the manuscripts which have benefited from them
most.
Of the value of other sorting methods, in particular statistical
methods, I remain sceptical. A numerical table of significant
agreements between every two manuscripts, as described on p. 38,
will provide objective confirmation of groupings suggested by
casual inspection, and will indicate how clear-cut they are (e.g.
how much more often G H I K L agree with each other than with
other manuscripts); but simply collecting the evidence, without
reducing it to figures, will probably have given a clear enough
picture already. Indeed, where groups of three or more manuscripts
are concerned it will have given a clearer one, for the information
that A agrees with 81 times, with C 92 times, and A with C 79
times does not enable us to deduce that ABC all agree together
73 times, or even once.
A more elaborate way of using such a table has recently been
advocated by J. G. Griffith 12 ). It involves comparing manuscripts
12
) J. G. Griffith: Taxonomic Study of the Manuscript Tradition of Juvenal,
Museum Helveticum 25, 1968, 101-138; Numerical Taxonomy and some primary

46

in respect of the whole numerical series of each one's agreements

with the others; no distinction is made between primary and se
condary readings, though agreements that may represent coin
cidental innovation are excluded. Suppose manuscript A shows the
following numbers of agreements with the others:
B F G H J
K L N O P R T U V Z
50 44 61 49 52 48 62 62 48 42 44 53 61 47 4 3 .
For each manuscript a similar series of numbers is found. The one
whose series matches that of most closely is then grouped most
closely with A. Finally the manuscripts are all arranged in a
'spectrum': those with the most dissimilar patterns of agreement
appear at opposite ends, with a continuous gradation from one
end to the other, while certain clusters or 'taxa' mark themselves
off along the line 13 ). The trouble with this kind of analysis is
that it is not clear what useful conclusions can be drawn from it.
Two manuscripts may be grouped together just because they show
no particular tendency to agree with any manuscript more than
any other, in other words because they are equally promiscuous,
even if they have no special similarity with each other textually u ) .
In some cases it is evident that the taxa reflect real affinity-groups,
in others it is not. In any case we are given no guidance as to the
distribution of ancient readings.

3. Diagnosis
When the evidence of the various sources for the text has been
collected and organized, apographa eliminated, hyparchetypes
and archetypes reconstructed where possible, and so on, the time
manuscripts of the Gospels, Journal of Theological Studies n. s. 20, 1969, 389-406.
He is preparing a book.
13
) For the method of performing these operations see the author's articles.
14
) An example is the taxon JU which Griffith constitutes for Juvenal.
47

has come to try to establish what the author originally wrote.

Sometimes this is a matter of choosing between transmitted
variants, sometimes it is a matter of going beyond them and
emending the text by conjecture, or adopting an emendation
already proposed. We will consider these alternatives separately;
but the requirements which a satisfactory solution must fulfil are
the same in both cases.
1. It must correspond in sense to what the author intended to
say, so far as this can be determined from the context.
2. It must correspond in language, style, and any relevant technical
points (metre, prose rhythm, avoidance of hiatus, etc.) to a way
in which the author might naturally have expressed that sense.
3. It must be fully compatible with the fact that the surviving
sources givre what they do; in other words it must be clear how
the presumed original reading could have been corrupted into
any different reading that is transmitted.
The fulfilment of these three conditions does not logically
guarantee that the true solution has been found, and there may
sometimes be more than one solution that fulfils them. An element
of uncertainty may therefore persist even if a reading is open to
no criticism - just as it may exist in places where the sources are
unanimous and what they offer unexceptionable. But often a
reading seems so exactly right that those most familiar with the
author can feel absolute certainty about it.

The evaluation of variants

Many variants are obviously wrong because they offend against
grammar, metre or the plain sense of the passage. The more carefully a critic has studied these things, naturally, the more such
faults he will detect; though it is possible to go too far, and to
fall into error by applying more rigid canons of language or logic
than the author observed.
If there are two or more variants in a given passage, and only
48

one of them is not obviously wrong, it does not follow that that one
is right. It may be a plausible but nevertheless incorrect emenda
tion by a scribe whose exemplar gave the corrupt text. And there
are other possibilities. That a plausible reading is not necessarily
a genuine one is proved by many thousands of places where more
than one of the variants is plausible. (In only a tiny minority of
these can authors' variants be involved.) By what criteria can we
judge between them?
Such variants are 'plausible' in that they satisfy the first two
of the three requirements. It remains, then, to consider how far
they satisfy the third. To some extent this is bound up with an
evaluation of the sources that attest them and with the inter
relationships of these sources. If the source of one of the variants
is a quotation, the assumption that inaccurate memory is responsible
for it may in many cases be the likeliest explanation of the discre
pancy. If you think that in general a reading which most of the
manuscripts give is more likely to be right than one which only
a few give, then to make the same assumption in a particular case
where the manuscripts are unevenly divided between plausible
readings, and to prefer the one attested by the majority, will be
to make the choice that (on your view) fits the facts best. That
would, of course, be a very naive principle. If the manuscripts
happen to be related as in the stemma on p. 32, it is easy to see
that a reading given only by has just as much chance of being
right as one in which BCDEF all agree.
M a n u s c r i p t s m u s t be w e i g h e d , n o t c o u n t e d . That is an
old slogan, one of several which deserve remembrance and com
ment in this connexion. A in the stemma just mentioned 'weighs'
equal to the other five copies combined. 'Weight' is not determined
solely by stemmatic considerations. Let us take a contaminated
tradition for which no stemma can be set up. Where the credit of a
plausible reading is concerned, a tenth-century manuscript whose
scribe is not given to emendation obviously carries greater weight
than a fifteenth-century one that is rich in copyist's conjectures, at
least if the reading might be a conjecture.
49

This is not to say that the age of a manuscript is necessarily a

guide to its quality. R e c e n t i o r e s , n o n d e t e r i o r e s : that is the
famous heading of a chapter in which Pasquali protested against
the tendency to equate the two terms, and showed that true read
ings are sometimes preserved only among the latest manuscripts *).
A propensity to emendation, so far from discrediting a manuscript,
may be symptomatic of an interest in the text that also prompted
the consultation of out-of-the-way copies, like the use of by
the late manuscript A in the stemma on page 44. Conversely,
very old copies such as papyri sometimes disappoint expectation
by giving a worse text than the medieval tradition instead of a
better one. The quality of a manuscript can only be established
by reading it. And when an opinion has been formed on the quality
of a manuscript, it can be used as a criterion only when other
criteria give no clear answer. The absurdity of following what
ever is regarded as the best manuscript so long as its readings
are not impossible is perhaps most clearly, and certainly most
entertainingly, exposed by Housman, D. Iunii Iuuenalis Saturae
(Cambridge 1905; 1931), pp. xi-xvi 2 ). Each variant must be judged
on its merits as a reading before the balance can be drawn and a
collective verdict passed. Since the collective judgment is entirely
derived from the individual judgments, it cannot be a ground
for modifying any of them, but only a ground for making a judg
ment where none could be made before. As Housman puts it,
"since we have found the most trustworthy MS in places where
its fidelity can be tested, we infer that it is also the most trust
worthy in places where no test can be applied... In thus commit
ting ourselves to the guidance of the best MS we cherish no hope
that it will always lead us right: we know that it will often lead us
wrong; but we know that any other MS would lead us wrong still
oftener."
In any case, merely to correlate good readings with good manu1

) S t o n a . . . , pp. 43-108.
) See also Frankel, Einleitung..., pp. 131-4.

50

scripts is a very poor way to satisfy the third requirement of a

convincing textual decision. It is not enough just to say "this
corruption of what I take to be the right reading is explained by
the fact that the manuscripts in which it appears are generally
corrupt manuscripts". A more particular explanation is called for,
in terms of known processes of textual change. Hence the criterion
u t r u m in a l t e r u m a b i t u r u m e r a t ? Which reading was the
more liable to be corrupted into the other? For example, if part
of the tradition gives a word or phrase which is absent in the
other part, and both versions give equally good sense and style,
one may ask whether it is something that a scribe might have
added, or whether it is easier to assume an omission. Where purely
visual errors with no psychological side to them are concerned,
the criterion has little applicability; it is mainly of use where there
has been some mental lapse, or a more or less conscious alteration.
Since the normal tendency is to simplify, to trivialize, to eliminate
the unfamiliar word or construction, the rule is praestat difficilior
lectio*). For instance, in Horace Odes 1,3,37, nil mortalibus ardui
est, some manuscripts have arduum, which is equally good Latin,
but also more everyday: to any copyist nil arduum would have
seemed more natural and obvious than nil ardui, and it is far more
likely that an original genitive was changed to the accusative
(whether deliberately or not) than vice versa.
When we choose the 'more difficult' reading, however, we must
be sure that it is in itself a plausible reading. The principle should
not be used in support of dubious syntax, or phrasing that it would
not have been natural for the author to use. There is an important
difference between a more difficult reading and a more unlikely
reading.
In deciding that one reading is derived from another and therefore to be eliminated, we are doing something similar to what we
3

) The principle was clearly enunciated by Clericus, Ars Cntica (Amsterdam 1696),
11. 293. For earlier hints of it see S. Timpanaro, La Genesi del mctodo del Lachmann
(Firenze 1963), p. 21 n. 1.
51

do when we decide that one manuscript is derived from another.

T h e principle can be extended. If there are m o r e than t w o variants
at a given place, we should try to put t h e m into a stemmatic rela
tionship (if this has not already been d o n e for the manuscripts in
which they appear). F o r instance, at Aristophanes Ach. 121
, there are three different readings
in the manuscripts, logically related as follows:

This does not mean

only that it derived
had the same stage
'stemma variantium'

(R, Suda)

()

().

that derived its text from A at that place,

it directly or indirectly from a copy which
of c o r r u p t i o n as we see in A. At 408 the
reads:

' ( R ) =

(Suda)

()

I
().
This assumes that the substitution of - for - t o o k place inde
pendently in the Suda (or an antecedent copy) and in A (or an ante
cedent copy). But a valid alternative w o u l d be

52

(R)

(Suda)

()

(),

with - being restored by conjecture at the last stage. This case

where more than one analysis is possible is not exceptional, and
the critic must consider what different hypotheses are available,
for they may lead to different choices for reading as the original
one.
Emendation
As the comparison of manuscripts may lead to the
of a lost archetype, so comparison of the variants
place may lead one to postulate another reading as
source. E.g. Hipponax fr. 104,49 (ap. Athenaeum

reconstruction
at a particular
their common
370a):

(Schneidewin)

()

(recentiores).

The reading of A represents a banalization of the Ionic into the

familiar form of the name of the festival, the other variant a
misreading of uncial as . Schneidewin's emendation accounts
for both readings and at the same time restores what Hipponax
meant to say in the correct dialect; it thus satisfies perfectly the
three requirements formulated on p. 48.
But the archetypal reading, reconstructed or extant, may be
unsatisfactory. In that case, further conjecture is called for, just
as it may be called for if there is complete agreement among the
manuscripts. It starts, so far as possible, from the 'paradosis'
(), which is a rather imprecise but convenient term
meaning 'the data furnished by the transmission, reduced to
essentials'. It would be almost true to define it as the text of the
archetype in a closed tradition, and the effective consensus of the
manuscripts (disregarding trivial or derivative variants) in an open
one. But reduction to essentials implies something further, namely
the elimination from the archetype-text or the consensus-text of
53

those features which we know, from our general knowledge of

the history of books and writing, to have been introduced since
the time of the author. The category includes orthographical
modernizations, capital letters, word division, punctuation and
other lectional signs.
To take orthography first, in Lucretius 4,1011, where the
primary manuscripts give
porro bominum mentes, magnis que motibus edunt
magna,
we must choose between quae (recentiores) and qui (Lachmann);
but as que is really only a way of writing quae, it would be legitimate
to describe quae as the paradosis and not an emendation. In
Semonides 7,4 the primary manuscripts give ,,, and a
Renaissance copy ., which is a form better attested for
early Ionic. From one point of view the paradosis may be said to
be .. But when one reflects that Semonides would have
written the contraction of simply as l, it appears that the
paradosis really amounts to an ambiguous ^: ,
. is some later person's interpretation of that, but we are free
to prefer the alternative interpretation 4 ). It is an emendation in
the sense that it corrects a presumed error, but not in the sense
that it postulates a form of the text for which evidence is lacking.
In ancient books there was no distinction of proper names, and
hardly any word division. Punctuation existed from at least the
fourth century B.C. and accents from the second, but the use of
these and other signs (such as the apostrophe marking an elision)
was very sporadic; in theory an accent or a breathing in a medieval
copy of a post-Hellenistic writer might go back to the author's
autograph, but in general all such features of the tradition will
4

) In the case of the Homeric poems, however, such decisions will have been made
by men familiar with the sound of the verse as preserved by generations of reciters,
and mistakes are much rarer than has sometimes been thought. See Glotta 44, 1967,
135-6.
54

represent some later person's interpretation of a text consisting of

virtually nothing but a continuous sequence of letters. The critic is
at liberty to re-interpret e. g. as , as , maxima
meque as maximam aeque.filiam artis as JHlia Martis, or to repunctuate,
even if he has taken a vow never to depart from the paradosis 5 ).
The same applies to the division and attribution of speeches in
dialogue texts. In ancient books a change of speaker was normally
signalled only by a dicolon(:) and/or a paragraphus, a dash over
the beginning of the first complete line. It is not certain whether
even this practice goes back to the earliest times, and the divisions
given by manuscripts are so often erroneous that they cannot be
regarded as useful evidence of the author's intentions 6 ). Certainly
the attribution of a speech to such-and-such an interlocutor rests
on no tradition that reaches back to the author (except perhaps
where the speaker makes his first appearance) but only on later
interpretation. The practice of regularly identifying the speaker
seems to have been invented by Theodoretus in the fifth century 7 ).
The critic is free to distribute the dialogue as best fits the sense.
In what circumstances is it legitimate to depart from the para
dosis, to entertain a conjecture? Many scholars would answer,
"only when it is clear that the paradosis cannot be right". Those
are scholars who will dismiss a conjecture from consideration on
the ground that it is 'unnecessary'. But it does not have to be
'necessary' in order to be true; and what we should be concerned
with is whether or nor it may be true. Consider Euripides, Hippolytus 99-101.

) Where a vox nihili (nonsense-word) is transmitted, however, accents etc. may be

valuable clues to what lies behind it, since they must have been supplied when the
text was in a more intelligible state.
6
) See J. Andncu, Le Dialogue antique (Pans 1954), pp. 288ff , J. C. B. Lowe,
Bulletin of the Institute of Classical Studies 9, 1962, 27-42, and good modern
editions of Menander (Lloyd-Jones's Dvscolus, Kassel's Sicyonius, Austin's Aspis
and Samia).
7
) N. G. Wilson, Classical Quarterly, n. s. 20, 1970, 305.

- ' ;
- '; , .
- ' .
So the medieval paradosis (with a variant ); but in a papyrus
of the third century B.C. the third line ends ]GAAC, doubtless
in place of the proper name. This is almost certainly the
right reading: the intrusion of the name to clarify a circumlocution
is a very familiar phenomenon (above, p. 23), whereas there is no
reason why an original should have been corrupted into
8 ). Before the papyrus appeared, however, anyone who had
suggested reading "e.g. " would have been told, "your
conjecture is unnecessary: is perfectly satisfactory". He
would have been justified in replying, "I am not saying that
is unsatisfactory, I am only warning you not to rely on it
too much, because this is just the sort of sentence in which a
proper name is liable to be interpolated". His warning would have
been timely and his conjecture correct. Probably no editor would
have thought it worth mentioning. Yet it would have deserved
mention, because it fulfilled the three requirements stated at the
beginning of this chapter, being in full accord with the sense and
with Euripides' style and metre, and easily compatible with the
fact that the paradosis gives .
This may seem to be opening the door to innumerable profitless
speculations. If we are to attend to every conjecture that is pos
sible, it may be said, there will be no end to it. But this is really not
so. The number of conjectures that genuinely satisfy the require
ments will not be large; and those that do ought to be attended to.
The critic should not be content to exercise his art only on passages
where his predecessors have exercised it. He should scrutinize
every single word of the text, asking himself whether it is in

) Barrett's argument for (. 439 of his edition) is answered by Merkelbach,

Zeitschrift fur Papyrologie und Epigraphik 1, 1967, 100.

56

keeping with the author's thought and expression 9 ), whether

there are other better or equally good ways of interpreting the
paradosis (e.g. a different punctuation), and whether the assumption that the wording of the paradosis reproduces what the author
wrote is the only hypothesis that satisfactorily accounts for it. We
want to know not only where the paradosis is certainly at fault, but
also how far we can depend on it in other places, and what the limits
of uncertainty are. The discovery of new sources (especially papyri)
has often revealed the presence of corruption where no one had
suspected it. It follows that one ought to be more suspicious.
The textual critic is a pathologist. It is his business to identify
disorders known to him from professional experience and from
textbooks (and the more he can supplement the latter from the
former, the more sagacious he will be). When he notices that all
is not well with a passage, however the paradosis is interpreted,
his first problem is to discover as precisely as possible where the
corruption lies. It may be obvious that one particular word is
wrong and everything else in good order, or it may not. In that
case he must go over the passage word by word, giving careful
thought to the meaning and to the author's writing habits, and
making preliminary decisions of the form "whatever has gone
wrong here, this word at least is just right and not to be tampered
with".
Finding the exact location of the corruption will sometimes
lead him at once to recognize its nature, and perhaps to see the
solution. At other times he will only be able to say where the
corruption is but not what kind it is; or what kind it is, but not
what exactly lies behind it. For instance, if something essential to
the syntax of a sentence or to the progress of an argument is missing, he may be able to say "there is a lacuna at this point, but
there is no knowing what it contained"; or "there is a lacuna
which must have contained the w o r d s . . . " ; or again "there is a
!)
) It is a good plan to make a translation. Nothing more effectively brings one face
to face with the difficulties of the text.

57

lacuna: the required sense would be given, for example, b y . . . " .

In this case the supplement proposed would fall into the category
of 'diagnostic' conjecture, that is, a conjecture which, while no
one can feel confident that it is right, serves the purpose of indi
cating the kind of sense that is really required or the kind of
corruption that may have occurred. If someone had conjectured
for in the Hippolytus line discussed above, that would
have been a diagnostic conjecture, by which was diagnosed
as an interpolation of a well-known kind 1 0 ). There is a dictum of
Haupt, quoted with approval by Housman and others: "If the
sense requires it, I am prepared to write Constantinopolitanus where
the manuscripts have the monosyllabic interjection 0". The point
he is making is that emendation must start from the sense. But the
failure to explain how Constantinopolitanus came to be corrupted
into may leave others with certain doubts as to whether that is
really what the sense was. Until those doubts are stilled, the
conjecture has the status of a diagnostic one. The vast majority
of corruptions in manuscripts are explicable. A conjecture which
presupposes an inexplicable corruption is not necessarily false,
but it is not fully convincing. The more completely the critic can
demonstrate that it satisfies the three requirements, the more
plausible it will seem. So far as the first requirement is concerned,
he can do this by analysing the argument of the passage, pointing
out defects in other interpretations, and comparing similar pas
sages from elsewhere. For the second, too, he will adduce evidence
about the author's practice generally and that of other authors of
the same period or genre. For the third, he will, if the corruption
is not of a widely-known sort, quote examples of similar ones.
When he claims that one word is a gloss on another, he will if
possible reinforce his case by showing from scholia or lexica that
the other word was so glossed. But he will be well advised not to
make his case too complex by assuming chancy multi-stage cor
ruptions, and not to rely too much on intricate palaeographical
) The concept of the diagnostic conjecture is due to Maas, Textkntik, p. 33.
58

arguments. These are the commonest faults in twentieth-century

emendation 11 ).
Even commoner is to deny the need for emendation and to
defend the paradosis at all costs. If good arguments can be produced
to show that the conjecture is mistaken (not merely 'unnecessary'),
that is fine. Understanding has advanced. All too often, however,
the defender only succeeds in showing that he has no feeling for
style, or does not know where to draw the line between the unusual
and the impossible; he asks "could these words bear the required
meaning ?" instead of "would the meaning have been expressed in
these words?" Sometimes one sees a conjecture dismissed simply
on the ground that all the manuscripts agree in a different reading.
As if they could not agree in a false reading, and as if it were not
in the very nature of a conjecture that it departs from them! Sometimes the emender must hold hinself back and admit that the means
to a solution are lacking: nescire quaedam magna pars sapientiae est
(Grotius). But to maintain that emendation generally is an idle
pursuit with little chance of success would be absurd. Hundreds
of conjectures have been confirmed (or at least raised to the status
of variants) by the appearance of papyri or other new sources 12 ).
Our knowledge of Greek and Latin, of the authors who wrote in
Greek and Latin, their ideas, styles, metres etc., and of the processes
of textual change, is not so inexact that we are helpless when our
manuscripts let us down.

) The palaeographical criterion is looked up to as an ideal by many whose understanding of palaeography is minimal, and who think that in order to make a conjecture palaeographically plausible it is only necessary to print it and the transmitted reading in capitals.
12
) Some Latin examples are collected by Havet, Manuel..., pp. 17-20. He points out
that in some cases the same conjecture may have suggested itself to an ancient
scribe as to the modern scholar; but these cases are certainly'in a minority.
59

PART II

EDITING A TEXT

1. Preparation
Is y o u r e d i t i o n r e a l l y n e c e s s a r y ? That is the first question.
Sometimes a new edition may be called for simply because no
existing one is easily available to a certain sector of the public schoolchildren, Poles, or scholars at large. If it is not a question of
filling some such gap, a new edition can only be justified if it
represents a marked advance on its predecessors in some respect,
whether in the fullness, accuracy or clarity with which the evidence for the text is presented, or in the judiciousness with which
it is used in constituting the text. The intending editor must
therefore be clear, first of all, that he is able to contribute something for which the critical world will be grateful. All too often
editions of classical authors appear that are not only no better
but distinctly worse than existing editions. Sometimes this is
due to carelessness in reporting the evidence or in correcting the
printer's proofs. The commonest cause, however, is lack of competence in fundamental matters such as language, style and metre.
Metre at least is reduced to rules: one would suppose that any
editor of a verse text would make a point of mastering the rules
relevant to his work, but in fact they frequently fail to (particularly
in the more southerly countries of Europe). Nor are they greatly
abashed when their mistakes are pointed out. They seem to feel
they have merely overlooked a minor technicality, and not to
realize that there is a large body of competent scholars whose
contempt is earned by nothing more surely than by metrical
61

blunders. Since bad editors are clearly for the most part quite
unaware of their limitations, it is difficult to offer advice that is
likely to deter them. But it may be worth pointing out a common
fallacy concerning the qualifications required. For editing a text
it is not a sufficient qualification to have a long-standing interest
in it, to have written articles or books about it, in short, to be
firmly associated with it in the public's mind. Nor even to have
investigated all the manuscripts and sketched the history of the
tradition: codicology and textual criticism are very different things,
and an expert on manuscripts may produce a dismal edition.
Publishers are sometimes at fault here. Wishing to publish an
edition of such-and-such an author to fill a place in some series,
they turn to whoever is known to have busied himself with that
author - no matter how - and invite him to undertake the task.
Flattered by this compliment, and sharing the publisher's assumption that his acquaintance with the text qualifies him to edit it, he
readily accedes, not stopping to reflect that this will expose his
philological weaknesses to his contemporaries and to posterity
more ruthlessly than anything else. A better policy for publishers,
when they want a good edition of something, would be to look for
someone who has done a good edition of something else, even if
he has not hitherto concerned himself with what they want 1 ).
Collecting the material
The editor's work begins with a period of study of what has
already been achieved by others. He does not necessarily read at
this stage everything that anyone has ever written on his author, but
he works through the main editions carefully, and whatever else
has been published on the manuscripts and other sources for the
text. In the absence of a complete and up-to-date special bibliography on his author, he will derive most help from library catalogues and from the following:
l

) Reviewers of critical editions should be chosen on the same principle.

62

W. Engelmann & . Preuss, Bibliotheca Scriptorum Classicorum, 8th ed. (Leipzig 1880). Covers literature from 1700 to 1878.
R. Klussmann, Bibliotheca Scriptorum Classicorum Graecorum
et Latinorum, = Bursians Jahresbericht (see below), Suppl.-Bde.
146. 151. 156. 165. Covers literature from 1878 to 1896.
S. Lambrino, Bibliographie de l'antiquite classique 1896-1914
(Paris 1951).
J. Marouzeau, Dix annees de bibliographie classique: 19141924
(Paris 1927-8).
J. Marouzeau and others, L'Annee philologique. Published re
gularly since 1928, covering literature from 1924 on.
(Bursians) Jahresbericht iiber die Fortschritte der cl. Altertumswissenschaft (Berlin & Gottingen 1875-1955); continued as
Lustrum (Gottingen, since 1957).
Gnomon (Berlin & Munchen, since 1925): Bibliographische
Beilage several times a year.
N. I. Herescu, Bibliographie de la litterature latine (Paris 1943).
Association Internationale d'etudes patristiques: Bulletin d'information et de liaison (Amsterdam, since 1968). Records pro
jected editions of patristic writers.
If earlier scholars have argued for certain relationships among
the manuscripts used by them, he should check their conclusions
as far as he can from the evidence available to him. He may find
that he cannot do so without fuller evidence about those manu
scripts; or he may suspect that there are other manuscripts which
have not so far been used at all. Then it is time to start doing some
collating for himself. Even if he is satisfied that all the manu
scripts have been investigated and their relationships correctly
assessed, he will be well advised to make his own collations of
the important ones, for two reasons. Firstly, it is very likely that
no complete collations have been published, only selected variants,
and he will want to make his own selection from the complete
evidence. Secondly, no one ever checks anybody else's collations
(or his own, for that matter) without finding mistakes in them.
Even what appears to be a very detailed collation is liable to
63

contain amazing mis-statements; and when it comes to making

inferences from its silence, the scope for error is large indeed.
He should not be afraid of collating because he has never done
it before, or because manuscript facsimiles that he has seen strike
him at first sight as indecipherable. Reading manuscripts is something that has to be learned, but it is by no means as difficult as
it may look to the uninitiated. Becoming an expert palaeographer,
able to date and identify hands, is another matter; but the main
thing, what one cannot easily get someone else to do for one, is
to be able to read them. There is great need to extend our knowledge
of classical manuscripts. People often assume that the task of collating has by and large been done, but there are many major
authors for whom dozens of manuscripts remain unread. The
sonner they are read, the better. Numerous manuscripts have been
lost since the Renaissance: how many of those that seem safe in
libraries today will still be there when another five hundred years
have passed?
Of the whole collating project, the hardest part to carry out
with complete success is probably the business of finding out
what manuscripts there are. For most of the libraries that come
into question, catalogues of manuscripts have been published in
book form or in periodicals. As far as Greek manuscripts are concerned we now have an excellent guide to these catalogues in
M. Richard, Repertoire des bibliotheques et des catalogues de
manuscrits grecs, 2nd ed. (Paris 1958), with the supplement to it
published in 1964. When he has consulted as many of the catalogues as he can, the inquirer may be recommended to apply to
the Institut de Recherche et d'Histoire des Textes, 15 quai
Anatole-France, Paris Vll e , where he may obtain help in supplementing his list of manuscripts (though he must not expect them
to do all his work for him). Papyri come in a different category: here
he can get his information from R. Pack, The Greek and Latin
Literary Texts from Greco-Roman Elgypt, 2nd ed. (Ann Arbor
1965), supplementing it from the papyrological bibliographies
published regularly in Aegyptus.
64

He should note down from the catalogues such information as

datings, identifications of scribes, the pages on which the work that
concerns him begins and ends, and the other works contained in
each manuscript. This last item may be a useful hint of a manuscript's affinities, for groupings of works changed frequently in
the Middle Ages and Renaissance. The investigator will not put
off the question of the interrelationships of the manuscripts till
he has finished collating them: he will be considering it while
he collates them, forming and modifying hypotheses all the time.
This will not only make the work considerably more interesting
to do (which will make him more alert and accurate while doing
it), it will also shorten it, as will be explained presently.
To do the collating he can either go to where the manuscript
is and ask to see it (some libraries will require a letter of recommendation from an official-looking source), or obtain photographic reproductions. (In most cases writing to the library will
produce the desired result; in cases of difficulty the Institut de
Recherche (see above) may be of assistance.) Both methods have
their advantages. By having the book in one's hands one is better
able to appreciate external features of its format and to distinguish different hands that have made corrections; at a difficult
place one can vary the angle of illumination as one likes, and
be sure that one has as good a view of what is to be seen as it
is possible to have. But occasionally things become clearer in
photographs; and they have the very important advantage that
one can easily refer to them again when some uncertainty arises
after the first collation. Going back to the library is much more
troublesome. Photographic copies may be broadly divided into
full-size reproductions of various types, and microfilms. The latter
are cheaper, and for most purposes perfectly adequate, but difficult to refer back to when one is away from a reading machine
or projector (though it can be done with a good magnifying-glass),
and for the same reason difficult to compare with one another.
It is possible, however, to inscribe any requisite reference numbers
on the margin of the film, using a fine pen and Indian ink.
65

The manuscript is compared with a printed edition word by

word, and the differences written down. Some people write them
in the margins of the edition, but even if the copy is interleaved
this does not give one room for more than a few manuscripts'
variants, and I usually use a separate notebook. It is essential in
this case to record in writing which edition has been used for
the collation, for if that is not known a collation loses much of
its value. (One must bear in mind the possibility that one's colla
tions will one day be used by someone else, and one must there
fore make sure that it is clear in this and in all other respects how
they are to be interpreted.) It is best to choose an edition which
is light to travel with, will always be easily available, and keeps
close to the paradosis (to minimize the amount of writing neces
sary); and to use the same one for each collation. Every effort
should be made to prevent confusion between the collations of
different manuscripts. If they are done into the printed copy, the
best thing is to use different coloured inks 2 ); in a notebook, the
manuscript should be identified at the top of every page. Care
must also be taken to avoid ambiguity about the location of the
variant. In prose texts the lines should be numbered down each
printed page and the numbers used for reference. If the variant
is for a word that comes twice in the same line, or might be read
as being for either of two similar words, it must be made clear
which one is in question.
If it is decided not to record certain orthographical trivialities
(e.g., in Greek, aspiration, or the presence or absence of sub
script or movable v), the fact should be stated. However, it is
advisable to record orthographical variants fairly systematically,
at least for portions of the text, for they can be of use (though not
by themselves) in working out the details of a stemma, and they
are not uninstructive in themselves. Corrections and marginal or
interlinear variants should be carefully recorded, with notes of
whether they are due to the original scribe or in another hand.
2

) Collations should always be in ink. If washable ink is used, beware of rain.

66

When collating in situ a manuscript that may be of some importance,

it is a good idea to note the point in the text at which each page
begins, for two reasons: one might then notice e.g. that an omission in another manuscript corresponded exactly to an opening
of this one (which might confirm indications that it was derived
from it); and if it is subsequently necessary to check the reading
in a certain passage, it is easy to order a photograph of the right
page.
It is useful to determine the manuscript's affinities if possible
before actually collating it. If they are not already known before
it is seen, they can often be quickly discovered with the help of
select lists of readings peculiar to the different manuscripts and
families. One or two such readings will prove nothing, but if (say)
ten passages from different parts of the text are looked up and
found to have the variants peculiar to a known branch of the tradition, it will be certain that a significant affinity has been found;
further comparisons will then reveal its nature more precisely.
If the manuscript is closely related to another that has already
been collated, its own collation can be done more quickly and also
more accurately by relating it to the other. One can write at the
start "Where the line-number alone is given, the reading is the
same as in Q", or "Has the same readings as Q except in the following places" 3 ). But it is wise to tick or underline the readings
in the collation of Q at the same time, as a precaution against later
doubts. This forces one to look specifically for each Q variant,
and it sometimes arouses suspicions - afterwards confirmed
that something in Q has been overlooked.
Should the manuscript turn out to be the exemplar from which
Q is derived, it will only be necessary (except in isolated places)
to underline those readings in the Q collation which it contains.
Conversely, if it turns out to be an apograph, it will only be
3

) If the relationship is known in advance, it is a great convenience to have the

two collations drawn up in parallel columns. The first will need to be wider than
the second.
67

necessary to note its additional errors and corrections. T h e r e will

be little point in m a k i n g a complete collation of an a p o g r a p h ;
there is some point, h o w e v e r , in collating a portion of the text,
to help determine its relationship to any other a p o g r a p h a that may
be discovered. T h e same applies to manuscripts deemed u n w o r t h y
of full collation for any other reason. T h e length of the text may
make it advisable to investigate the whole tradition in the first
instance on the basis of sample p o r t i o n s . If so, the p o r t i o n s to be
studied should be taken both from near the beginning and from
near the end, because it is not u n c o m m o n for a manuscript's
allegiances to change in the course of a w o r k .
So far I have been speaking only of the direct tradition. Ancillary sources t o o may call for research. An e p i t o m e or a translation
has its o w n manuscript tradition. If the q u o t a t i o n s have not been
systematically collected, that may be s o m e t h i n g else with which
progress can be made, by reading t h r o u g h the likely authors
or consulting indexes to them. If they have been collected, it will
still be necessary to look them up in the most up-to-date editions
of the q u o t i n g a u t h o r s , to verify the references and to see exactly
what the textual evidence from that source is. Where such editions
are suspected of being unreliable or founded on an inadequate
basis, it may be w o r t h the trouble to consult manuscripts of the
authors.

Digestion
T h e processes of analysing the relationships of the various sources
and evaluating the variant readings and conjectures have been
described in earlier chapters. T h e editor is n o w at the stage when
he can perform these operations in a m o r e definitive manner. He
is by n o w very familiar with his a u t h o r , and it is desirable that he
should be not much less familiar with any other authors w h o are
particularly relevant because they are imitated by his author, or
imitate him, or write in the same m a n n e r or on the same subjects.
He completes, as far as possible, his reading of other scholars'
68

interpretations of the text and discussions of its difficulties. (In the

case of major a u t h o r s this is frankly impossible, the quantity of the
secondary literature is t o o great. But o n e must d o what one can,
and try to pick out the grains from the chaff.) And then, in days
of unhurried c o n t e m p l a t i o n , preferably assisted by a w o r d index
or concordance, he decides what he is g o i n g to print in his text.
This involves m o r e than just deciding which are the true read
ings and which p r o b l e m s must be left unsettled. Careful t h o u g h t
should be given to p u n c t u a t i o n , which can be a great help or
hindrance to following the a u t h o r ' s train of ideas, and which is
of course entirely a matter for the editor's discretion. T h e n there
is the question of o r t h o g r a p h y . As a general rule it would seem
most rational to impose consistently the spelling that the original
author is most likely to have used (for which the manuscript
tradition may not be the best evidence). It is true that he himself
may have been inconsistent, and it may be argued that the best
manuscript authority should be followed on each occasion. But
this will be no reliable guide to his practice; we shall surely come
nearer the t r u t h by regularizing the spelling than by c o m m i t t i n g
ourselves to the vagaries of the tradition.
T h e general rule, h o w e v e r , is subject to qualifications. N o one
would welcome an edition of Aeschylus in which the Choephori
began
11 , ' ,
/ ' ,
and for early Greek generally one will use the standardized Ionic
alphabet, a l t h o u g h this sometimes means using different spellings
for sounds that were originally written the same, and the same
spelling for sounds that were originally written differently. In
Latin there is not the p r o b l e m of different alphabetic systems, but
notions of the correct way to spell things were m o r e fluid until
the first century of the E m p i r e , and here again ( t h o u g h with less
justification) the c o n v e n t i o n has been established of presenting
authors at least of the late Republic in the o r t h o g r a p h y of a some69

what later period. Late or vulgar texts raise other difficulties: it

is often impossible to distinguish between the barbarisms of
copyists and those of the original. In this situation, rather than
impose a consistent system which can only be chosen rather
arbitrarily, it is better to follow the paradosis, not under the
delusion that it is at all reliable, but as the most convenient way
of exhibiting it.
An associated problem that may face the editor is that of deciding
what exactly it is that he is trying to constitute. A book transmitted
to us may represent a re-working or rearrangement of older
material, or the end product of several re-workings, and the
editor must be clear which phase of its history he is restoring.
Convention is inconsistent. Editors of Greek tragedies are agreed
in trying to purge the texts of actors' interpolations, whereas
editors of Homer do not normally mark as spurious passages of
clearly secondary origin such as the Doloneia. The standard
edition of Stobaeus rightly aims to show each passage not as its
author wrote it but in the form in which the anthologist received
it. Editors of the Palatine Anthology do not try to restore the
arrangement of the older anthologies from which it depends. On
the other hand they try to print the original text of each epigram,
not the tenth-century text. These choices are sensible. They may
be said to be based on the two principles of seeking the useful
and not attempting the impossible.
In the case of a work that survives in more than one recension,
the editor must either give each recension separately or choose
one as a representative. He must not conflate them into a hybrid
version which never existed (though he may use one to correct
copyists' errors in another).
The use of computers
The possibility of using computers to help the editor in some of
his labours has been discussed by Dom J. Froger, La Critique des
textes et son automatisation (Paris 1968), chapter 5. It appears
70

that the time has not yet come when manuscripts can be collated
automatically; machines have not yet been devised which can
cope with the variations inherent in handwriting. If provided
with suitably prepared transcriptions of the manuscripts, purged
of coincidental errors, a computer could draw up a clumsy and
unselective critical apparatus; and it could in principle - where
there was no contamination! - work out an 'unoriented' stemma.
That means, supposing that six manuscripts were related as shown
on p. 32, that it could work out a scheme

A
-B
simply by comparing the variants, without regard to whether they
were right or wrong; but this scheme would be capable of suspen
sion from any point, e.g.

(
D

71

The correct orientation could only be determined by evaluating

the quality of the variants, which no machine is capable of doing.
Since only a minority of textual traditions are closed, and these
easily analysed by ordinary human wit, the very considerable
trouble involved in submitting them to a computer does not appear
worth while. At present, it seems, computers can serve us best by
making concordances and the more unsubtle kinds of metrical
analysis 4 ).

2. Presentation
Prefatory material
It is the editor who is mainly responsible for the layout of his
book, and he should take pains to arrange it as conveniently as
possible for the reader - not only the reader who works through
it from cover to cover, but also the one who only needs to consult
it briefly, who is not deeply familiar with the text and its transmission, and wants to extract information quickly and easily.
Wishing to know what construction to put on what he finds
in the critical apparatus, such a reader is likely to turn to the
introduction, which should be so set out, with section- and pageheadings, that he finds at once where the sources for the text
are discussed. He may want to see what is said about a particular
manuscript, and he should be guided to the place by some signpost:
a separate paragraph-heading, bold type in the text, or, best of
all, the siglum printed in the margin. He should then find the essential information; he is most likely to be interested in the date
and general character of the manuscript, and its affinities. If it is
found more convenient to discuss affinities after the account of
the individual manuscripts, this too should be clearly signposted.
') On the use of computers in questions of style and authorship see K. J. Dover,
Lysias and the Corpus Lysiacum (Berkeley & Los Angeles 1968), chapter VI;
B. Fischer, Journal of Theological Studies, n.s. 21, 1970, 297-304.
72

If the conclusions of the discussion can be shown in the form of

a stemma, a stemma should be printed; nothing makes them easier
to locate or to comprehend. Quotations may not need to be discussed in the introduction, but sections should be devoted to
sources such as scholia, epitomes, and translations. If the introduction is long, a list of contents is useful.
If a bibliography is provided, it should aim above all to inform
the reader about the work of those editors and other scholars
whose contribution is significant enough to have earned them a
place in the critical apparatus. Ideally, whenever the reader finds
a scholar named in the apparatus as having proposed an emendation or defended the paradosis, he should have the means to
identify the relevant publication, so that if he wishes he can consult it and read the scholar's own argument. (In practice the editor
must sometimes cite conjectures, mentioned by a previous editor,
whose provenance he is unable to discover.) Publications that are
only occasionally of importance can be specified in the apparatus
itself; l ) those to which more constant reference is made are better
listed separately. Editions are usually listed in chronological order,
but other works should be arranged in alphabetical order of
authors. If several books or articles by the same scholar are to be
recorded, it is a good idea to number them, and the reference in
the apparatus can then take the concise form * Meyer 3 ' or * Meyer 3
p. 268'.
Immediately before the text 2 ) the list of manuscript sigla should
be found, together with the explanation of any other unusual
symbols or abbreviations used in the text or apparatus. It is a
convenience if the manuscripts' dates are mentioned again here
as well as their identities, and also their groupings, the limits of
their content (if they do not contain the whole text), and page1

) The abbreviations 'I.e.', 'op.cit.' are to be avoided unless the work has been
named immediately before.
2
) 'The text' here includes ancient prefatory matter, Hypotheses, lists of characters
etc.
73

references to the discussions of them in the introduction. For an

example see Bethe's edition of Pollux, where however two criti
cisms can be made: the list could be clearer typographically, and
it should have been printed in both volumes of text.

Choice of sigla
Manuscript sigla in current use should not be changed unless
there is something particularly confusing about them (for instance,
if the same manuscript has been given different sigla in different
works of the same author). Where new sigla are necessary, the
following principles can be observed with advantage.
1. For individua 1 extant manuscripts, and only for these, use capital
letters of the Latin alphabet. If there are not enough letters
(though there should be, if the tradition has been analysed
properly), one may resort to Greek capitals, or to e.g. Aa, Ab,
or AjA 2 ; it is best to avoid superior figures (A1 A 2 ), because
they are commonly used for distinguishing hands in the same
manuscript. In editions of Latin authors it is advisable to distin
guish the siglum visually from the adjacent variant; this can
be done by printing it in bold type, but a better typographical
effect is obtained by using italic capitals. In Greek apparatuses
either upright or sloping capitals can be used, only not both: to
use A and A to mean different things would be to guarantee
confusion and error. Letters with a mnemonic value should be
chosen where possible, e.g. = Monacensis; A = the most
notable manuscript; = Triclinius' copy. It is not necessary to
assign sigla to manuscripts that are only cited in a few places.
2. Fragments of ancient copies, whether papyrus, parchment or
ostracon, are often given sigla like II, II 6 , 25, 4 1 , which
conveniently draw attention to their antiquity. The use of
superior figures here is familiar enough to be acceptable, in
spite of what is said above, and aesthetically it is preferable to
full-size figures. It has the disadvantage, though, that other
74

qualifications such as 'ac', '' (see below, p. 93) cannot very

satisfactorily be appended at the superior level.
3. For manuscript families or reconstructed hyparchetypes 3 ), use
lower-case Greek or Latin letters. It is best to use Greek letters
in editions of Latin texts and vice versa. If Greek ones are used
in Greek texts, they should be set from a different type fount
from that used for the variants themselves (as in Pfeiffer's
apparatus to Callimachus , Hymns). Latin letters should be in
italic, or, if not, bold.
The symbol or or is often used to mean 'all the manu
scripts' ; this is better avoided, however, if other letters of the same
fount are being used for other purposes. The symbol ~ (originally
standing for 'Stephanus') is often used, especially in Latin edi
tions, to mean 'one or more late manuscripts'.
4. Any sources of a different order should be represented by
symbols or abbreviations of a different order. Thus the use of
S for the Suda in the Bude Aristophanes is not very satisfactory
beside R, V, etc.; Su would have been better. is often used for
'scholia'.
The body of the edition: general layout
The text will occupy the upper part of the page. Where it is recon
structed from excerpts given by different sources, or where dif
ferent recensions or versions in different languages have to be
presented, they are in most cases most clearly exhibited in parallel
columns or on facing pages. (Examples: W. H. Roscher, Die
Hippokratische Schrift von der Siebenzahl; H. Diels, Doxographi
Graeci; Damascius, Vita Isidori, ed. Zintzen.) Where this is
unsuitable, they should be printed consecutively for each indepen
dent block of text, as e.g. in Hausrath's edition of the Aesopic
Fables, where three recensions are printed, the page is not divided
3
) The editor should make it clear which he means: whether a means ', B, C
and D ' or 'the exemplar from which ABCD are derived'. It makes a difference.

75

into three columns or horizontal layers, but each fable is given

first in one version, then in the second, then in the third. With a
more continuous text such as the Life of Aesop, it is better to
print the recensions as separate wholes (as in Perry's Aesopica)
than to interlace them chapter by chapter.
Below the text on each page will follow, in this order: any
registers other than the critical apparatus (testimonia, etc., see
below); the critical apparatus; and commentary or translation if
these are to appear on the same page as the text. (The best place for
a translation, however, is facing the text. As for a commentary,
although there is clearly some advantage in having text and note
on the same page, there is still greater inconvenience if the effect
is to reduce the amount of text on each page to a few lines. If the
bulk of the edition justifies printing the commentary in a separate
volume, that is the handiest arrangement.)
With editions of fragments an alternative layout is possible,
with the apparatus to each fragment following it immediately,
before the text of the next fragment. (P^xample: D. L. Page, Poetae
Melici Graeci.) This is quite unobjectionable so long as the fragments are short, but if they run over the page, and even more if
they run over two pages, it becomes inconvenient, and the editor
would be better advised to follow the normal arrangement, although it means a more complicated job for the printer.
The margins will be used for numeration. Sometimes they are
also convenient for indications of manuscript attestation (below,
p. 83), or of the sources of a compilatory work such as the Suda
(see Adler's edition). The heading at the top of the page should
be informative. If the volume contains more than one author or
work, the reader must be able to see at once from the page-heading
which one he has opened at. If a work is divided up in units that
are liable to exceed a page in length - books, long chapters or
fragments, groupings by subject or metre, years (in annalistic
historians) - he again needs help from the page-heading. The
left- and right-hand pages can be used to give different grades of
information, and in addition section-numbers can be presented
76

as a 'shoulder head' at the inner corner. Here is an example from

the OCT Hesiod, pp. 126-7:
(left-hand page)
I (right-hand page)
sive
[28-30 30-31]
AKOLIDAE
(title of poem)
|
(section)
(fragment numbers)
The pages should be numbered.
Text
Both prose and verse should be printed in numbered lines. In
verse texts the numbering will be continuous from the beginning
of the poem, book or fragment; in prose texts it should start
afresh at the top of each page and run to the bottom, except where
the pages and lineation of an older edition have become established
as the means of reference (as in Plato and Aristotle). Experience
has shown that this is the most convenient way of correlating
text and apparatus. The numbering should be by fives 4 ). In the
case of verse texts it will be the main means of reference, and
it is best printed on the outer margins (i.e. to the left of the text
on the left-hand page and to the right on the right), or else to the
left of the text throughout (where there will be no interference
from long lines; the numbers will be a constant distance from
the verbiage). If an alternative numeration is to be printed too
which should only be done if it enjoys some currency - it should
be in brackets or in distinctly smaller type. In prose texts the linea4

) It is sometimes claimed that numbering every third or fourth line, instead of

every fifth, makes it easier to find a reference. I disagree, believing that it is quicker
to find one's way from the simple stake-points 60 65 70 75 80 than from the
more complex scries 60 63 66 69 72 75 78 81 or 60 64 68 72 76 80. One does not
need to think so hard about the individual numbers.
For the case where the accepted numeration reflects an obsolete colometry, see
Barrett, Hunpides: Hippolytos, p. 94.

tion will normally not be the common means of reference, and it

will go on the inner margins, unless the outer ones are fairly free
of other numbers. The more prominent position and type will be
reserved for the conventional chapter- and section-numbers.
Numbers should not have to be sought in the middle of a line;
even if a new section begins there, the number should be in the
margin. (The exact point of transition can be marked with a divider
if there is any ambiguity.) Arabic numerals should always be preferred to Roman or Greek, except in the numbering of books (or
of columns, in papyrus texts); where Greek numerals are traditional for that purpose, they should be accompanied by their
Roman or Arabic equivalents. Care should be taken that numerals
of different orders of significance are well distinguished typographically. Where it is necessary to note the pagination of an
older edition as well as chapters and sections, it is helpful to add
an initial; e.g. in K. Nickau's edition of 'Ammonius', '44 Va.' in
the margin explains itself at once, whereas the figure by itself
would have left one unsure and made one look somewhere else for
clarification.
Established numerations should be retained as far as possible.
If the editor decides that the usual division of paragraphs is
unsatisfactory, he can change it without moving the numbers. If
his study of the manuscripts has resulted in the discovery of additional sections or verses, he should give them numbers like '53a'
which do not disturb the rest of the series. Snell's editions of
Pindar's fragments give an illustration of how an inherited
numeration can sometimes be adapted to accommodate new discoveries. It must be admitted that new numerations are sometimes
necessary, but all too often they are made for frivolous reasons.
The editor should ask himself whether a new numeration is really
going to be convenient for the user, bearing in mind that he will
continue to encounter references to the older system. If it is, it
should be made as simple as possible.
Literary quotations and allusions in the text, including selfquotations ('as I have written elsewhere'), should be identified.
78

The references can be given in one of the registers below the

text, or, perhaps more conveniently if the quotations are short
and not too numerous, between brackets in the text itself. The
same applies to dates given by an ancient author in Olympiads
etc., which should be furnished with the equivalent in our reckon
ing.
Quotations should be presented in the form in which the quoting
author gave them, so far as this can be determined, not adjusted
to what we believe the quoted author wrote. Verse quotations
should be printed as verse, unless the quoting author has destroyed
the metre. Otherwise quotations should be distinguished as such,
insofar as they are verbatim, by inverted commas or spaced type.
Spaced type is particularly suitable for picking out verbatim ele
ments in a loose paraphrase: see, for an example, Plato Protagoras
339-346 in the OCT edition.
Inverted commas (double ones in Greek texts) should also be
used for speeches, except in those texts which, like drama, consist
wholly of dialogue with no narrative framework. Here their place
is taken by abbreviated speakers' names. These are normally put
at the beginning of the line, but when there is a change of speaker
within a verse one may adopt either of the following arrangements:
(a) . .
, ;
.
' .
. ;
.
' ' >\
.
;
(b) (.) .
, ; (.) ' .
(.) ; (.) ' ' \ (.) ;
The second may lead to lines so long that the printer has to break
them anyway (which should be done at a change of speaker), but
it saves space and makes line-references easy to find.
79

The following critical symbols have a place in the text. (They

can of course also be used in the critical apparatus as convenient.)
( ) Besides being used for ordinary parentheses, round
brackets are used for the expansion of abbreviations, e.g.
M(arcus) Cicero sialutetri) d(ixif) Ser(uio) Sulpicio.
< ) Angle brackets enclose letters, words or passages added
to the transmitted text by conjecture or from a parallel source.
They can also be used with a blank space or *** or metrical
symbols between them, to indicate the editor's belief that some
thing has been omitted in the course of transmission (or the
asterisks can be used alone). They should not be used to mark
letters which an emender has substituted for something else; thus
qui(ay should signify that the transmitted reading is qui, and
where quia is an emendation of quid it should be printed without
brackets 5 ).
[ ] Square brackets have commonly been used for editorial
deletions. But among papyrologists and epigraphists it is now
firmly established practice to use them to mark off parts of the
text lost through physical damage to the extant source; and since
in practice no sharp line can be drawn between texts edited from
papyri and other texts, it is highly desirable that square brackets
be reserved for that purpose. When the number of letters missing
can be estimated, it is indicated by the corresponding number of
dots below the level of the line, [. . .] 6 ), or by a figure, [-16-].
When it cannot be estimated, print [] or [***], or, if the
distance between the brackets indicates the size of the gap, [ ].
{ } Braces replace square brackets as the sign of editorial
deletion. They can be used in combination with angle brackets
to show that a transposition has been made, e.g. ( OTL)
{} , but as this involves
printing the transposed element twice it soon becomes cumber5
) In some editions it would be printed quitf. This rather ugly practice is now
outmoded.
e
) Grouping these dots in fives makes them easier to count; see e.g. MerkelbachWest, Fragmenta Hesiodea. Metrical symbols can be printed above the dots.

80

some if mmr than a word or short phrase is in question. Trans

positions of verses should be shown simply by the line-numbers,
as e.g. in the OCT Hesiod, pp. 5, 14, 23, 31. (Transposed lines
should never be renumbered.)
[ ] Double brackets enclose letters or words that a scribe has
deleted in the manuscript itself. If such letters can no longer be
read, use dots as above, |[. . .]. The symbol |||, repeated for each
letter, is also used for successful erasures.
These signs are used by papyrologists to enclose insertions
by a scribe after he has made his original line. One could represent
a scribal alteration of to by [] ', but it is more elegant to
print simply and note in the apparatus '8 ante corr.'; if the
reader fails to consult the apparatus, that is his fault.
L j Half brackets are a logical modification of full square
brackets. In papyrus texts they indicate that the papyrus itself is
broken or worn away but that the supplement is supplied by
another source and is not conjectural. In other texts, by extension,
they can be used to show the absence of a particular source,
whether because it is damaged or because it has a shortened ver
sion of the text. They might well be used, for instance, in a text of
Nonnus' Dionysiaca to show which letters are preserved in the
Berlin papyrus, or in one of Athenaeus to show how much is
attested by the epitome. Complications, arise, however, if there is
more than one of these intermittent sources. can be used for
a second one, and Bethe's Pollux shows how more elaborate in
formation can be conveyed; but it may be wondered whether such
feats of typography are often worth while, especially when they
endanger the legibility of the text.
f f Obeli mark words which the editor judges to be corrupt.
If only one word is suspect, only one obelus is needed: suhsidiis
magnis \epicuri constabilitas. If the editor cannot limit the corrup
tion to one word, he places his two obeli so as to define the area
within which it is to be sought: declinare quis est qui \possit cernere
sese\.
Dots under letters indicate that they are uncertainly deci81

phered 7 ). They are mainly used for papyri and inscriptions, but
there is no reason not to use them generally. They are not easily
combined with subscript iotas, so adscript iotas should be used
in texts where dotted letters are needed.
The use of metrical signs to guide the reader where there may
be ambiguity is not to be scorned. SnelPs editions of Pindar and
Bacchylides are a model in this respect. Besides giving the metrical
scheme at the beginning of each song, he helps us to read without
constantly consulting it, by printing e.g. /, ,
, , .8). Some editors of Plautus
and Terence print ictus-marks, and they might well go further
in signalling unclassical prosody 9 ).
Between the text and the apparatus
There are some kinds of information that are best presented in
a separate register or registers above the critical apparatus.
Scholia, in those cases where they are sparse enough to be con
veniently printed with the text, should go immediately below it.
Then the sources for the text should be specified insofar as they
are variable. If different manuscripts are available for different
parts of the work, different poems in a collection, etc., the details
should be shown on each page in whatever is the most suitable
form for the circumstances.
Here are a few examples.
7
) When the trace cannot be identified at all and the space above the dot is blank,
one should insist on the dot remaining below the level of the line, to distinguish
it from a full stop.
H
) His use in the fragments of the symbol (corresponding to the ancient coronis)
to mark the beginning or end of a song is also commendable.
}) Why should we not, indeed, revive for classical Latin texts the apex (') with
which the Romans themselves, for about three centuries from the age of Sulla,
found it convenient to mark long vowels? We would not use it for every long
vowel (nor did they), but it would be very useful for forms like ablative natura,
accusative plural ciuisy and for advertizing hidden quantities in such words as iwxy
dixit.

82

(Pindar, ed. Snell) V, B E F G H = v ; E F G H = , G H = . Printed

in the critical apparatus at the b e g i n n i n g of the ode. T h e value of
the collective symbols and the separate status of V are con
veyed to save the reader t u r n i n g to the prefatory pages.
(Eur. H i p p . , ed. Barrett) c o d d . : (446-59 ) V (469-74 H)
C D E L. Printed a b o v e the apparatus page by page. H K are available
for the passages stated, the rest for the whole page. T h e manuscript
g r o u p i n g s are indicated by spacing.
(Menandri Sententiae, ed. Jakel) 864 || 865 || 866 || 867 ||
868 || 869-870 U || 871 || 8 7 2 || 873-875 || 876-877 .
( A m m o n i u s , ed. Nickau) 1-8 o m . || 9-12 o m . Mp || 13-14 om. ||
15-17 o m . .
(Petronius, ed. Bucheler) L, L O , etc., printed in the margin of
the text at the t o p of each page and w h e n e v e r the attestation
changes. This is a satisfactory alternative to the position below
the text p r o v i d e d that only a small n u m b e r of sources have to be
named. D r a c h m a n n uses it in his edition of the scholia to P i n d a r ;
it is more usual in editions of scholia to specify the manuscripts
at the end of each scholium or alternative version thereof, as in
Schwartz's edition of the Euripides scholia. See p. 98.
Q u o t a t i o n s by later writers should be specified, and allusions
or imitations at least w h e r e they provide evidence of the text read.
Allusions etc. should be distinguished by 'cf.' or 'respicit', 'imit.'.
If one of the writers mentioned is d e p e n d e n t u p o n another, this
should be remarked (see e.g. Pfeiffer at Callim. H y m n . 3,180).
Some editors give n o t only the reference but an extract from the
context in which the q u o t a t i o n appears (e.g. Rzach in his big
edition of Hesiod), and this is often a help in assessing its value
for the text. But it is also possible to indicate the reason for the
q u o t a t i o n , where it matters, m u c h m o r e briefly. Here, for example,
is a t e s t i m o n i u m on Pindar Ol. 2,45 (,
) in the full and in a shorter form.
(a) 45 Et. G e n . 47,7 Cal. = M a g n . 18,48 "
, , '
, " ".
83

(b) 45 \\8 (-?.- disertim) : lit. Gen. 47,7 Cal. = Magn.

18,48.
However, as the reader's attention will have to be drawn to the
unmetrical variant in the apparatus if at all, all that is really neces
sary is: 45 \ \ 8 . 0. lit. Gen. 47,7 Cal. = Magn. 18,48. In the appara
tus he will find something like: 45 - codd., Ktym. disertim:
corr. Tricl.
It sometimes happens that a piece of text extending over several
pages is copied out by a later writer. It is helpful to the reader in
such a case to give the appropriate reference on each new page.
The same applies to the converse situation, where the author
being edited has copied out a long passage from an earlier writer
(who therebv becomes relevant to the constitution of the text in
the same way as a quoter); and similarly where he has reproduced
the substance of such a passage in his own words, and again where
he and another writer are evidently following the same lost
source 10 ). Such parallel texts should be distinguished from direct
quotations by 'cf.', at least if there is any danger of ambiguity: in
some authors there will not be. The decision whether to print
the references to sources, parallels, imitations and quotations in
separate registers must likewise be governed by the particular case.
Rzach's big Hesiod is a successful example of elaboration, with
its four separate registers below the text: Homeri loci similes,
Poetarum (ceterorum) imitationes ct loci similes, Testes (i.e.
quotations and allusions), Varia lectio. By contrast, Maass's
Aratus shows how sources, parallels and tcstimonia can all be
noted in the same register without unclarity, with the help of the
sign 8 and 'cf.' Some editors go further and incorporate them in
the critical apparatus, to the detriment of its perspicuity; this
procedure cannot be recommended.
Before discussing the particular problems of the critical appara
tus, I should deal with certain aspects of layout which are common
,0
) For an example of the technique see Mras's edition of Fuscbius, Praeparatio
evangelica.

84

to it and the o t h e r registers. Reference to the text is made by speci

fying the line of verse, or in prose texts the line of the page (except
in cases like P l a t o ) 1 1 ) . If several items refer to the same line, the
numeral is not repeated. Sometimes further precision is called for.
Suppose a q u o t a t i o n only covers part of a line; then one must p u t :
636 (-) Ht. M. s. v. 08 (meaning line 636 as far as );
554 (.) - 555 (8.) Ht. . s. v. >. In the critical apparatus
the information to be conveyed will usually be sufficient for the
identification of the w o r d or phrase in q u e s t i o n ; where it is not,
brackets may again be used, e.g. 1096 () L, or a colon, 1096
: L. square bracket has most often been used for this pur
pose, 1096 ] L, but as it is sometimes necessary to use square
brackets for r e p o r t i n g readings from papyri and other damaged
manuscripts, it may be better to avoid that. If occurs twice in
the same line, one must be sure to make clear which is m e a n t :
1096 p r i u s : or ( 1 ).
Separate, n o n - o v e r l a p p i n g items are parted from each other by
a broad space (as in the O C T series) or by the divider-sign || (as
in the Bude series; similarly in m o d e r n T e u b n e r editions except
that a single vertical stroke is used between entries relating to
the same line). Spaces have the theoretical drawback that they
disappear when the second item begins at the b e g i n n i n g of a line
of type, t h o u g h in practice confusion seldom arises, and can be
avoided by care at the proof-reading stage. O v e r l a p p i n g items in
the critical apparatus should be treated in the same way, where
the points at issue arc unconnected (e.g. 717 -^
c o d d . : - Stob. B e r g k ) ; overlapping
items in the quotations-register are perhaps better linked, e.g.
371-2 + 374 sch. Pind. O . 7,72; (-) A m m o n . s. v. ,
Hust. in Horn. 1527,57; 371-2 ( - ') sch. E u r . T r o . 8 5 5 ;
1

') The line-number is often printed in hold type, but it stands out perfectly well
in ordinary tvpc, as users of the OCT \olumes can see. When the numbers run
in more than one series, the series-number must be in bold and the line-number
m light face, see the OCT editions of Hesiod (fragments), Plato, Aristotle.
85

371 + 374 sch. Pind. I. 5,1; (-) sch. Eur. Ph. 175; 371 sch.
A. R. 4,54.
Items in the apparatus that are physically separate in their re
ference but interdependent should normally be brought together,
e.g. Soph. Tr. 1021-2 . . . Musgrave: . . .
codd., rather than 1021 Musgrave: codd. 1022
Musgrave: codd.

The critical apparatus

Critical apparatuses have more than one use. The most essential
one is to inform the reader which parts of the printed text depend
on emendation and which parts are subject to uncertainty. But
apparatuses are also what most people depend on for instruction
about the character of particular manuscripts and scribes, and of
manuscripts and scribes generally. Unfortunately, the more fully
an apparatus caters for the latter need, the less handy it is for the
former; the important variants have to be discerned amid crowds
of unimportant ones 1 2 ). The editor must decide what principle
he is going to follow, and select his material accordingly.
few basic rules can be laid down:
1. The readings of apographa and other manuscripts which seem
to contain nothing of independent value should be omitted,
except where the exemplar is illegible or where an interesting
emendation is involved.
2. Variants of a merely orthographical nature should be omitted
unless they represent real alternatives (: ), or
unless manuscript evidence is relevant to the choice.
3. Worthless conjectures should be passed over in silence; more
precisely, conjectures which are not only unacceptable but also
12

) device sometimes employed is to relegate the less important ones to an

appendix; see e.g. Kenney's OCT edition of Ovid's Amores etc. One might also
pick out the main ones in larger or heavier type.
86

fail to suggest a new and plausible line of approach to the

problem 13 ).
4. Anything from any source (including scholarly conjecture)
that may either be or point towards the true reading should be
reported.
The apparatus should be in Latin, which has proved itself the
most convenient for the purpose (except for papyri and inscrip
tions, see p. 94). Names of scholars and periodicals should not
be latinized. In editions of Latin authors italic type is used except
for the variants themselves, the line-numbers and the punctuation.
The material is arranged on the following principles. Each entry
begins with a specification of the place in the text which is in
question, unless it is the same as for the preceding entry (see above,
p. 85). The reader is already given one reading by the text itself.
If it is not a conjecture, he can usually infer which sources attest it
by elimination of those quoted for other readings, so it is not
necessary to mention it in the apparatus (so long as it is clear which
word or words the entry refers to) 1 4 ). If it is included, it should
be put first, unless it is represented by the formula 'corr. Haupt'
(which has its place after the transmitted reading(s) and before
any further conjectures that are to be mentioned). Alternative
readings follow, in this order of precedence: direct manuscript
tradition; indirect tradition (testimonia etc.); conjectures in order
of merit. The editor ought to have a regular order to name the
manuscripts in 1 5 ), but he should depart from it whenever the
13

) Conjectures that have been confirmed e.g. by a papyrus deserve to be recorded

as such, for the honour of their authors and as evidence that emendation is a worth
while endeavour.
14
) An apparatus which regularly leaves the reading of the text to be understood
is called 'negative'. There is no need for the editor to make a firm decision between
the positive and negative apparatus; different treatments may be convenient in
different places. negative entry rather suggests an aberration, and I would
recommend using a positive one where the rejected variant is well attested or
judged worthy of consideration as a serious alternative.
15
) He will naturally group cognate ones together.
87

logical connexion between variants is better brought out by a

different ordering. Similarly he may find reason to couple a
quotation-variant with one of the manuscript readings, or to
place two conjectures together irrespective of the order of merit.
The variants and conjectures presented ought properly to fit
the same hole in the text. Thus the variants at Ar. Ach. 121 men
tioned on p. 52 must not be presented like this:
121 R Suda: :
That implies falsely that is omitted by . " "
should have been written. However, strict adherence to this rule
would sometimes involve excessive repetition, and I confess that
1 sometimes break it if I think there is no possibility of confusion,
particularly when reporting conjectures, as at Archil. 122,4.
( ) , Mahly: Valckenaer:
/ Bentley, etc. Minor variations on a reading can be given in
brackets:
Bergk ( Welcker, Boissonade):
A etc.
Or, if they are of similar status, they can be separated by a mere
comma instead of the colon which normally separates alternative
readings and expresses their opposition to each other 1 6 ). Thus
' , ' : ' C
indicates that the choice is really between two alternatives ( or
), not three. In a negative entry, where the choice is not really
between the readings in the apparatus at all, it might be better to
avoid colons and to use commas or semicolons.
16

) There is some variety of usage on this matter. Modern Tcubner editions do

not use colons; the Bude series uses them only after the first reading, the one in
the text; similarly the Corpus Paravianum, but with commas after the second
and subsequent readings, thus:
539 Metanira Hetnsius: melanira Li, metania IG 1 , menaha 0, menalca D etc.
The question is not important, but the system that I recommend is the most
flexible and expressive.
88

A transmitted reading should be quoted in the form in which

it appears in the source, obviously. But this rule too is subject to
qualifications. It is not necessary or customary to print the readings
of pre-minuscule manuscripts in capitals. (Ziegler's edition of
Cicero, De re publica, however, reproduces the uncial script of
the Vatican palimpsest to pleasing effect. This kind of fidelity is
particularly helpful when corrupt Greek appears in a Latin text,
cf. p. 27) One will write " II 1 8 codd." even though the
papyrus has no accent or breathing; if its reading is being given
by itself, on the other hand, one might as well be exact and write
" 11 18<t . Abbreviations in manuscripts need not be reproduced
unless they are ambiguous or help to explain the origin of another
variant. Sometimes there is good reason for the editor to make
an abbreviation of his own: in reporting variations of word order,
aequo animo ferre nemo T: nemo ae. an. f. E: ae. an. n. f. 8;
in dealing with long words,
: - D,
which, besides saving space, focuses attention on the variant
element; and to avoid making a statement about part of a word
in which unimportant variants exist. For example, suppose the
whole truth is
affirmasse A: adfirmasse BE: affirmauisse C: adfirmauisse D,
one may save space and at the same time clarify the two issues
by printing
aff- AC: adf- BDE

-asse ABE: -auisse CD,

or simply the second divergence if the first is not thought worth

reporting. But however insignificant the first is, it should not
be disregarded to the extent of printing
affirmasse ABE: -auisse CD,
which involves a positive mis-statement. It would be better to put
affirmasse {vel adf-) ABE: etc.
89

Also to be avoided is
affirmasse A: adf- BE: -auisse CD,
where it is not clear whether CD have aff- or adf-.
In general, abbreviation of readings should be kept within modest
bounds. Too much of it will cause the reader bother.
The same applies to abbreviation in the editor's own Latin,
though familiar abbreviations like om., add., ci., transp. are preferable to their full forms. Scholars' names should be abbreviated
sparingly; it is all right when they are long and famous (Wil.)
or of frequent occurrence in the particular apparatus (that to
Quintus of Smyrna, for instance, is full of Rhodomann and Zimmermann, who have a good claim to be shortened), but the casual
user of the edition does not want to have to turn to the list of sigla
for an explanation of Bk. or Hu.11). Especially to be deprecated
is the use of abbreviations which do not suggest a human being
at all, for instance c = Cunaeus, g = Graefe. The incautious consuiter of the edition will certainly think that readings so labelled
have some sort of manuscript authority, and he may not feel he
has anything to gain by foraging in the introduction for codicological details 18 ).
The basic information to be given about each reading is its
source. This is done in the ordinary case simply by placing the
appropriate sigla or name after the reading 19 ). If the editor wants
to emphasize his confidence in a conjecture which he has adopted,
17
) The misleading practice of omitting a full stop with abbreviated names
('Dalec' = Dalecampius) should be eschewed.
18
) Scholars' initials can also be mistaken for manuscript sigla. In my apparatus
to Theognis I have written O t t o Schneider* to avoid confusion on the one hand
with J. G. Schneider and on the other with the manuscript O.
19
) The statement of sources of a transmitted reading should not be augmented
by names of editors or critics who have approved it. Similarly with conjectures,
only the original propounder should be named, with the place of publication if
necessary (see p. 73). There is no point in inserting 'ci.' (coniecit) except where
an emendation has to be distinguished from a decipherment of a difficult manuscript.

90

he can transpose "uinxerunt Heinsius (or scripsi): iunxerunt coddV

into "iunxerunt codd.\ corr. Heinsius (or correxi, or emendavi)".
(The difference of meaning between 'corr? or *em? and V/.' deserves
more respect than many editors give it.) Omissions, additions,
etc. are shown thus:
(i) Omission by a source.
672 om. b (i.e. the whole verse is omitted).
672 deest in b (carries less suggestion that b is at fault).
11 ab exitio urbium om. L (or deest in L).
(ii) Expunction by a scholar.
11 qui omnis hominis scit nomen del. Rumpe/sti/^cben.
If the deletion has been marked in the text by brackets, all that is
needed is
11 qui - nomen del. Rumpelstil^cbeny or
11 { } Rumpelstil^cben.
'11 del. Rumpelstil%chen\ however, may be misleading, at least in a
verse text, since it gives the impression that a whole line has
been condemned.
(iii) Extra words in a source.
9 post aliquando verba scilicet post
mortuorum habent MN.

resurrectionem

(iv) Words added by conjecture 20 ).

25 natam/>0j7 fortunatam add. (or rest, or suppl.) Gandalf,
or
25 fortunatam <natam) Gandalf.
If a lacuna is marked in the text by < ) , all that is needed is
25 ^natam) Gandalf.

20
) Conjecture by the copyist of a manuscript will be treated in the same way as
conjecture by a modern critic.

91

If (natam) is accepted in the text, all that is needed is

25 suppl. Candalf, or {

) Candalf.

(v) Transpositions.
26-8 illud tibi - non uenerint post p. 31 JO nostra tramp.
Bartsch
1 arma canoque uirum Tilgenwit^.
Where the transposition is adopted in the text:
10-12 illud tibi - non uenerint ante p. 30,28 quaedam
babent coda1.: tramp. Bartsch.
1 uirumque cano codd.: tramp, liigenwit^.
Where the fact of transposition is shown in the text by the versenumbers or by the combination of<( ) and { }, all that is needed
is
213-4 transp. Hermann
16 transp. Koraes.
If a verse has been transferred some way from its transmitted
position, the reader will notice a gap in the numeration without
immediately being able to see the reason for it. This calls for a
note in the apparatus such as 136 v. post 1158'. (A note is also
called for if the numeration is discontinuous for any other reason,
e.g. after Catullus 17.)
Mention may here be made of a more laconic style of apparatus
favoured in particular by Wilamowitz. Here is a specimen from
the Choephon.
(Text)
900 Tj)\) 8
906-7 bracketed
908 8
915

(Apparatus)
: Auratus (i.e. cod.: corr.
Auratus)
906,7 del. Berlage
Auratus:
*Y/ok:Wil

(He might have written : Auratus in 908, but felt that the
reader might then hesitate before deciding that was the word
92

in the text that had been substituted for .) The system has not
won much acceptance, and it seems that most people prefer things
to be a little more explicit.
The choice of sigla has been discussed on p. 74. Often a quali
fication is needed, such as 'before correction', 'over an erasure',
'by a second hand'. Here one may use an abbreviated Latin phrase
in ordinary type on the line: ante corr. (or a. c), in ras(ura), m. sec.
(or rec.) \ or more compressed compendia at the superior level:
/Vlc, /Vr, A 2 . (With a papyrus called II 9 , however, one will have to
revert to MI9 m. r e c ' or 9 m 2 '.) These must be explained in the
list of sigla. Care must be taken to forestall confusion over the
referents of these qualifications; the comma is the simplest way
of making clear what is to be taken together. For instance, instead
of . D' write , . D' or ., D', whichever is meant;
instead of 'AD a c G' write 'A a c D a c G' or 'AG,D a c ' 2 1 ).
Superior type is also useful when there is variation between
different manuscripts of an indirect source. At Theognis 724
for is given by and by cod. A of Stobaeus.
' Stob. A' might well lead to confusion, since the most
important manuscript of Theognis is also known as A. 'p, cod. A
Stobaei' is safer but cumbrous; 'p Stob. A ' is better than either.
To embark on discussion of the merits of variants will rapidly
enlarge and obscure the apparatus. But when a telling point can
be made with a couple of words or mention of a parallel, there
is every advantage in making it. To point out, for example, that
a certain omission is explained by homoeoteleuton may save the
reader from the temptation to attach some greater significance
to it, and the editor himself from the need to refer to it in his
commentary. To indicate the reason for a conjecture may be to
avoid mystification or impetuous scorn. The apparatus is not
unsuitable for an interpretation of a difficult phrase, either, even
if there is to be no mention of an alternative reading. The resolu21
) Commas or spaces can also be used to remind the reader of the manuscript
groupings.

93

tion of an obscurity is a contribution to the examination of the

soundness of the text.

Some special types of edition. Papyri, inscriptions

Editions of papyrus texts and inscriptions fall into t w o classes
which may be described as 'scientific' and 'literary'. T h e first is
particularly a p p r o p r i a t e to the first publication of a new text
(which should be accompanied by a p h o t o g r a p h ) or to a revision
carried out on the original. It reproduces the formal layout of the
original and conveys an accurate picture of its whole appearance.
It preserves its alphabet (without necessarily imitating its letter
forms) and records its spelling, its p u n c t u a t i o n and its lection
signs, marginal additions, etc., in the apparatus if not in the text.
M o d e r n reading signs may be introduced p r o v i d e d that they d o
not obscure ancient o n e s ; FpaTpa FaXsioic. Oapptv
may be printed w i t h o u t risk that anyone will
suppose the inscription to be furnished with accents and long
signs, but with a papyrus that has some accents the simplest way
to c o m m u n i c a t e the details is to print them as they are. W h e r e
there are difficulties of decipherment or interpretation, the m e t h o d
o f ' d i p l o m a t i c ' transcription may be r e c o m m e n d e d 2 2 ). complete
ly objective transcription that adds n o t h i n g to what is visible on
the original is printed together with an interpretation of it, e.g.
-[ ][
,- [
'-
[

|] [
, TTICTOV - [
ISp'Scadk [
' [

T h e apparatus includes a careful description of doubtful letters,

or an indication of the different possibilities. This goes beyond
the ordinary range of apparatus Latin, and it may be better if the
editor uses his o w n l a n g u a g e ; compare the English palaeographical
22

) See E. G. Turner, Greek Papvn (Oxford 1968), p. 71. The system might also
be used for a text which depends on one fairly corrupt medieval manuscript.

94

notes in recent volumes of the Oxyrhynchus Papyri with the Latin

ones in Lobel and Page, Poetarum Lesbiorum Fragmenta. Re
ference is by column and line. It is customary to use Roman
numerals for column-numbers.
In what I call literary edition of a papyrus text or inscription,
less attention is paid to the layout of the original. The text is pre
sented more as any other piece of prose or verse would be; there
is less emphasis on the physical copy and more on the composition
itself. The difference between this and the more interpretative
sort of 'scientific' edition is illustrated by the text of the first
Delphic Paean as given on p. 141 of PowelPs Collectanea Alexandrina, in comparison with that on pp. 142-4. The lines are redivided,
and melodic spellings like eliminated. This kind of
edition is suitable for a literary work such as Aristotle's Constitu
tion of Athens, or a play of Menander, where numeration by
chapter or verses takes the place of column and line 23 ). The ap
paratus too will be more like that to an ordinary text. There will
be no need to record details like accents in a papyrus, except
where they affect the interpretation. Anyone with a special interest
in them will naturally turn to the original, scientific publication,
to photographs, or to the papyrus itself.
The distinction between scientific and literary edition is of
course not absolute. Their characteristics can be blended in dif
ferent ways, and something between the two may be what is most
suitable in a particular case.
Fragment collections
Special problems are involved in editing the fragments of lost
works gathered from references by other authors. The first ques
tion is what to include. Some editions include only verbatim quo2:{

) So long as reference is by the lines of the scribe or stonemason, it is better to

reproduce this hncation than to have to mark it by a system of dividers, as is done
e.g. in Dittenberger's Sylloge Inscnptionum Graecarum: that makes it a slow
business to find a line-reference.
95

tations, but it is arbitrary to make this a principle; a statement

that the author told such-and-such a story, for example, may be
far more valuable than an uninformative verbatim fragmenr. If the
editor wishes to publish only selected fragments, let him make
his selection in some sensible way. If his collection is meant to
be complete, he must include testimonia - not biographical
statements about the author or aesthetic judgments on his work,
but everything that helps to compensate for the loss of the work
by supplying evidence about its form or contents.
The sources for each fragment must be specified, and in many
cases something of the context in which a quotation occurs must
be given in order for the reader to orient himself. The most
straightforward form of presentation is to print the fragment surrounded by its context (but picked out by larger or spaced type
as appropriate) and preceded by the source-reference. Examples:
Merkelbach-West, Fragmenta Hesiodea; Jacoby, Fragmente der
griechischen Historiker. If several sources give the fragment in
different contexts, it may be abbreviated after its first appearance,
as in Hes. fr. 205. Dependence of one source on another should
be indicated, as in Hes. frr. 62, 126, 170.
An alternative format often adopted consists of printing the
fragment in isolation and the sources and contexts somewhere
below. Examples: Diehl, Anthologia Lyrica Graeca; Pfeiffer,
Callimachus. This gives the page a tidier appearance, but makes
it harder to read the fragment in its context. It is useful as a variation on the first system when the fragment is not given complete
by any source but is a reconstruction from two or more; see e.g.
Archilochus fr. 5 or 43 in my Iambi et F^legi Graeci I. The latter
fragment will also serve to illustrate how different sizes of type
may be used to distinguish more important from less important
sources.
Fragment-numbers should be printed prominently above the
fragment or in the left-hand margin; in the apparatus they should
be in bold type to distinguish them from line-numbers, in the case
of verse fragments, but in an edition of prose fragments (unless
96

thev are very short) it is better to use the lineation of the page for
reference as in an ordinary prose text. If alternative numerations
are to be given, they may be added in brackets after the main
number in the text (as in Page, Poetae iMelici Graeci) or in the ap
paratus (as in Fragmenta Hesiodea), or reserved for a separate
table. There must also be a table for converting the old numeration
to the new. People will more often want to trace an old-style
reference in the new edition than vice versa.
For remarks on the position of the apparatus on the page, see
p. 76. In many cases it will be necessary to cite in the apparatus
manuscripts of numerous different authors. The more often the
reader can be given a brief note about their relationships, the
more intelligently he can use the apparatus. For examples see my
Iambi et Elegi Graeci I. ix-xi and the apparatus to Archil, frr. 115,
122, 129.
Scholia
Most bodies of scholia exist in different recensions. These should
be edited together, not in different volumes or parts of a volume,
nor on the other hand conflated into a hybrid text, but each
distinct version of each scholium consecutively. (Minor variations
of wording in individual manuscripts do not constitute a 'distinct
version'.) Each of these items should start on a new line. Unless
it begins with a linking-formula such as , it should be
prefaced with a lemma indicating which piece of text is the sub
ject of the comment. This may or may not be transmitted in the
manuscript(s). If it is not, it should be supplied by the editor in
brackets 24 ). long supplied lemma may be abbreviated, e.g.
Iliad 1,13-16 ( - ). The lemma should be printed
in bold or spaced type. It will itself be preceded by the reference
of the chapter, section or verse from which it comes, unless this
24

) Round brackets, as used by Drachmann in the Pindar scholia, are more ap

propriate than angular ones, since the supplement is of something understood
rather than omitted in error.

is the same as for the preceding item. Some editors distinguish by

a series of letters the several scholia included in the same reference,
whether or not they refer to the same lemma, e.g.:
Pind. Ol. 3,12 a. (:) . #
b. : .
c : . . B g (C),
DQ ( - ).
It is a good idea. The lines of the page will also be numbered in
the inner margin, as in a normal prose text, and this numeration
used in the apparatus.
The manuscript sources for each item are best stated at the end of
it, in the text, as shown in the example just given. The parenthesis
after DQ (which are marked off by the comma after (C)) informs
the reader that in those manuscripts the scholium is followed by
the one on verse 17; such dislocations are common, and though
they should be corrected, intelligibility sometimes depends on
their being recorded. The outer margin can be used for indications
of the scholium's origin in cases where it can be inferred. Example:
Erbse, Scholia Graeca in Homeri Iliadem.
Grammatical material in scholia, especially to Greek poets, is
often closely cognate with material in scholia to other authors or
in etymologica etc. The parallels should as far as possible be sought
out and cited in a register above the apparatus. Examples: Drachmann, Erbse, opp. citt.; Wendel, Scholia in Apollonium Rhodium.

Indexes
What kinds of index are required will depend on the nature of the
text and the degree to which satisfactory indexes have already
been published. The most usual sorts are the index of proper
names, the index of authors quoted or alluded to, the index of
subjects (in a work of a didactic nature), and the general index of
words. In an edition of fragments, besides the numeration-con98

cordance mentioned above, an index of the sources is useful papyri, quoting authors etc. I n d i c e s n o n s u n t p r a e t e r necess i t a t e m m u l t i p l i c a n d i . There is seldom any advantage and
often some disadvantage in separating the proper names from the
index of words, for instance; and while Drachmann deserves
benevolence for his anxiety to help the user of his Pindar scholia,
the user for his part needs some persistence if he is to find what he
wants amid the fifty-two alphabetical sequences at the end of the
edition. Page-headings should be used to make clear which index
each page-opening belongs to.
In an index of proper names, the entries are best given in the
same language as the text, and in the nominative case, unless some
special interest attaches to the case-forms. Different bearers of the
same name must be distinguished, and some closer identification
is always useful, e.g.
' ()

L. Arruntius (historicus,
cos. 22 a. Ch.)
Asellius(? Sabinus; cf.
(RE 34) 25 )
PIR 2 , A 1213)
' (, RE 7) (. Pomponius) Atticus.
If there are many references, it is a great help to the user to
classify them and give an indication of what each passage is about.
In the case of a prose work the editor should consider whether
references by page and line of the edition are not more convenient
for the reader than references by book, chapter and section. (It
should be said in favour of the latter that they can be used in con
junction with a different edition, to save turning pages backwards
and forwards while looking up a series of references.) If the re
ferences consist of more than one numeral, it will make for clarity
to separate adjacent references by a semicolon rather than a single
point. But '36,26; 36,29;' should be abbreviated as '36,26.29;' not
25
) I.e. the 34th holder of the name in Pauly-Wissowa. The examples are from the
OCT editions of Diogenes Laertius and Seneca's Letters.

99

'36.26;29;\ '1,2,1; 1,3,3; 1,3,5; 1,6,7;' mav be abbreviated as

1 2 , 1 ; 3,3.5; 6,7;'.
An index of authors can be combined with an index of names,
as in the OCT edition of Gellius. It should list the passages quoted
or alluded to in order, with a note of the edition used if necessary,
Epicurus ed. Usener
fr. 132: 57,4
133: 61,5
135: 58,6.
In an index of subjects the entries will usually be best left in the
language of the text. No one will find a Latin index to a Greek
work convenient to use. If there is a word index, the subject
index will naturally be incorporated in it.
word index may be selective or complete. There are limits to
the degree of completeness that is useful: there is no point in
listing every occurrence of , for example, since anyone making
a special study of would have to work through the whole
text in any case. The sensible procedure would be to pick out the
instances of in special uses, in combination with other particles,
etc. and to indicate that the entry has been limited to these. It is
obviously a great help to the user to distinguish between separate
meanings and constructions to some extent, e.g. between +
accusative and + genitive, or ut comparative and ut final.
It is very desirable in indexing some texts - those of literary or
linguistic and not merely technical interest - to distinguish the
various forms of nouns, verbs, etc. 26 ). Naturally they should all
26

) A difficulty arises when the index covers a group of authors or works, like
Fatouros' Index Verborum zur fruhgnechischen Lynk. Someone looking up (say)
the instances of in Anacreon has to pick out the Anacreon references from
those given for each case-form; someone looking up the instances in all the poets
has to keep jumping from edition to edition. The user's convenience is perhaps
best served in such a situation by abandoning the analysis of forms except in
entries where they are particularly significant.
100

be grouped together: under . The entry might

read simply
29,2. 3 bis; 30,1; 6,7.

But suppose the author also used forms of , . The

user of the index would look for forms of after those
entries. The entry should then take the form
() 29,2 etc.
Where there is uncertainty about the occurrence of a word or
form because of some textual problem, warning must be given by
means of some adjunct to the reference, \ci.)\ \v. /.)', or a symbol
such as an asterisk.
Printing
The printer should be offered a clear and well-spaced manuscript
or typescript on paper of uniform size 27) with the pages numbered
continuously throughout. Only one side of the paper should be
used. Any corrections that have been made should be very clearly
marked. Instructions about page-headings etc. should be provided.
Footnotes for the introduction may either be placed at the foot
of each sheet or on separate sheets. In either case they should be
numbered serially through the chapter or the whole. Similarly
with the critical apparatus and other registers to go below the
text; the editor will almost certainly find it convenient to set
them out on separate sheets. Unless he starts each item on a new
line, he will have to be careful to indicate the spaces between
them, especially at the end of a line.
Making a new copy of the text is laborious and involves the
danger of error, particularly errors of omission. It is common
practice to send the printer instead a copy of an existing edition
with the required changes marked (in ink) as if on printer's proofs.
27

) Except that for the text itself a corrected copy of a printed edition can be sub
mitted (see below), and for the indexes a pack of cards or slips.
101

This too has its dangers. It is a well-documented fact that errors

and misprints persist from edition to edition as a result of it 2 8 ),
and one can only advise that the editor takes the greatest care to
see that the one he uses correctly reproduces the paradosis in the
places where he does not choose to depart from the paradosis.
He must also see that it has been brought into conformity with
his wishes in matters such as punctuation, numeration, the use
of capitals.
Making alterations in proof causes extra delay and expense.
The editor should reduce the need for them as far as possible by
verifying references at the manuscript stage and generally seeing
that his manuscript is correct, unambiguous and consistent with
regard to abbreviations etc., and that it reflects his final and
settled opinions. But if he has grown wiser by the time the proofs
come, the printer's interests must yield to the reader's. The correction should be devised in such a way as to cause as little disturbance to the typesetting as possible, for instance by compensating
for a deletion with an insertion of similar length nearby. This is
particularly kind if the proofs are already arranged in the form
of pages, as opposed to galley-proofs 29 ).
Page-proofs are necessary, of course, before page- and linereferences can be adjusted. The line-numbers in the apparatus to
a prose text must be inserted or corrected at this stage, unless
the pagination of an older edition has been reproduced exactly.
Indexes in which references are given by page and line are best
not made till now.
Conclusion
The problems which different texts present, to the editor or to
the textual critic, vary enormously, and one must be flexible
28

) See A. Severyns, Texte et Apparat. Histoire critique d'une tradition imprimee

(Bruxelles 1962); Frankel, Flinleitung ..., p. 123.
29
) This does not seem the place for an account of the marks used in correcting
proofs. Conventions differ somewhat in different countries.
102

enough to follow whatever course is most suited to the particular

circumstances. The advice contained in the foregoing pages will
not meet every possible case, I am sure. But if the editor holds
fast to the ideals of accuracy, clarity and elegance, does what the
subject demands, and treats the reader as a deserving but not
necessarily patient friend, there is a good hope that his edition
will be welcomed just as warmly as his scholarship merits.

PART III

SPECIMEN PASSAGES

1. Hesiod, Theogony 176-200

For my first illustration I have chosen a text transmitted in a
fairly large number of manuscripts. I have complete collations
of thirty at my disposal, and partial collations of several more.
Many of them are of no individual importance, but in what follows
1 report their readings in full in order to give an idea of the extent
and character of manuscript variation in a tradition of this kind.
No stemma can be constructed, but the majority of manuscripts
fall into clearly defined groups. The hyparchetype a is represented
by four descendants, O V W X . None of them is older than the
14th century, but agreements with an 11th-century fragment and
with the perhaps pre-Eustathian allegorical commentary of Johannes Diaconus Galenus suggest that a may represent a kind of
Byzantine vulgate. Not dissimilar is r, a lost copy represented by
seven extant manuscripts and their apographa; three of the seven
come from a intermediate copy u which deserts r a little later in
the poem. After leaving r, u will follow k, which may have been
a recension of the Palaeologan era drawing on several sources.
In lines 176-200, k is represented only by one extant manuscript,
K. Of similar character to k is the hyparchetype b, represented
by five manuscripts. One of these, L, is 14th-century, the others
depend on a 15th-century model m.
S, the oldest manuscript available here (1280), is of Planudean
origin, and characterized by emendation as well as by eclecticism.
The same applies to Triclinius' autograph copy Tr, made a little
105

before 1320. T h e r e m a i n i n g m a n u s c r i p t s mostly b e l o n g to the

u n d i s t i n g u i s h e d family c, w h i c h has its heyday in the 15th century
b u t has a p r e c u r s o r in Q a b o u t 1300.
' ,
, '
* 6 '
, ,
180

, *
, '
. *
,
* '

185

' ,
, ' ' ,
' ' ' .
'
' ' ,

190

' ,
' * *
'
' , .
' ,

195

* '
{ }
, '
,
' , '

200

, .

176 Vat. 1948

A n accusative w o u l d be possible, b u t t h e correct form w o u l d be
, and a single m e m b e r of the c family is m o s t unlikely to
preserve a true r e a d i n g by itself.
177 (i) A r u n d e l 522 : u ( superscr. U 2 )
106

The variants are restricted to the r family, and seem to be

independent of each other. Neither has an obvious explanation;
possibly the scribe who wrote , seeing the words 'longing
for sex', at once thought of his own longing.
(ii) : Vat. 1948
The scribe was unfamiliar with the poetic word, - may be
an echo of - or an anticipation of , or, if the exemplar
was written in two columns, it may come from ^ at the end
of 175, which would have appeared immediately above, -
represents the substitution of a phonetically equivalent but com
moner ending, as if from a verb in -.
178 (i) K L S T r Par. 2772 Laur. 31,32, . 2 schol.
Et. Gen. : Mosq. 469 ( in ras.?), Aristonicus :
Q : OV r Par. 2833 Vrat. Rehd. 35, . L 1 : W :
2 ex : Mosq. 470 Phillipps 11723
Senensis I,IX,3, Par. 2834 : Vat. 1948 :
Paley : Heyne : , Ahrens
Scholia on this line and on Iliad 23,160 discuss the accentuation
of . In both, is commended but it is admitted that
manuscripts give . Aristonicus is named as the authority
for the correct form. It is clear, then, that was the ancient
reading, and although it is not found elsewhere, it is credible as
a by-form of , however we accent it. Those manuscripts which
have preserved it mostly give the accentuation which the scholia
describe as current. In a, r, and m it was displaced by the ordinary
form , to the detriment of the metre. But the copyist of one
of the m manuscripts, Mosq. 469, had the scholium before him
and was able to restore (with Aristonicus' accentuation).
The scholium may also be responsible for the accentuations
in W and in Vat. 1948. in four other c manu
scripts may represent either a misreading of or a metrical
emendation of . ( 2 ) is certainly the latter, inspired
by epic forms such as for . Q's springs from con
fusion with , which the context put into the scribe's mind.
107

It should not be taken as s u p p o r t for Ahrens's conjecture, which

was based on suspicion of the form / and on A n t i m a c h u s '
use of the a d v e r b / in connexion with the castration of
U r a n o s . T h e other conjectures also aim at being rid of /,
but retain the w o r d meaning hiding-place, which Hesiod used in
174 and is likely e n o u g h to have used again. Heyne's introduction
of links the clause with the following one more closely than
is usual in epic narrative. Paley assumes an original / with
the first syllable scanned long as with / in T h e o g n . 1099,
in H o m e r and H i p p o n a x , etc. / w o u l d then be an early
emendation of what seemed unmetrical. This is certainly an easier
change than that p r e s u p p o s e d by Ahrens, b u t still harder to believe
than that / is original.
(ii) ' T r
T h e particle here is indefensible. Triclinius inserted it on
metrical g r o u n d s w i t h o u t regard to its p r o p e r function, being
unfamiliar with the rare scansion .
(iii) A r u n d e l 522 Q Par. 2772 Laur. 31,32 Vat. Barb.
43 : V : M o s q . 470 Phillipps 11723 Senensis
I , I X , 3 Vat. 1948, Barb. 43 in marg.
in Q , three c and one r manuscript is phonetically equi
valent to for the Byzantine scribe. H e made the error
perhaps while mentally identifying the verb as ; he would
not have written for . presumably stood in
the c o m m o n ancestor of Q and c, but arose independently in the r
manuscript. O t h e r scribes failed to recognize the verb, seeing
instead other verbs that / might well a c c o m p a n y ; in a
s u b - g r o u p of c may be influenced by below.
179 (i) M o s q . 470 Senensis I , I X , 3 , Phillipps 11723 ante corr.
T h e same s u b - g r o u p is at fault. Perhaps a red initial was over
looked,
(ii) b : 2
is a correct form, but gives a less usual r h y t h m for this
place in the verse, and can be explained as an assimilation to
108

 in 159 and 173. The unmetrical correction in X is

probably also inspired by the recent evidence that is the
proper form of the adjective; but the -ov ending is retained.
(iii) a r Mosq. 469 Vrat. Rehd. 35 Q c Et. Gen.
The ordinary form has made extensive inroads in the tradition:
not, however, in metre-conscious copies such as S and Tr.
180 (i) S
Influenced by the preceding and perhaps the mascu
line associations of the following -.
(ii) Vat. 1948
The manuscript gave the same spelling in 175, and Hesychius
has * (for ). Simple assimilation of syl
lables will be responsible,
(iii) Arundel 522
Assimilation to the preceding accusatives.
(iv) hie et in 188 Nauck
This is the dialect form used by Hesiod in WD 512, where he
is speaking of animals. No manuscripts there have altered it to
, and it is not very plausible that it should have become
twice in the Theogony; nor is it justifiable to suppose that
in speaking of the gods Hesiod would not use the Homeric form.
181 (i) Vat. 1948 : Arundel 522
Only the adverb is used in the sense 'quickly' in early epic.
The scribe of Vat. 1948 may have been expecting a nominative
participle; - is after all a much commoner ending than
-. The compendium for is easily mistaken for , so that
- might be explained as a conflation of - with a correc
tion .
(ii) V Vat. 1332, U ante corr. : Casan. 356.
The first variant is merely orthographical. The second may
have been influenced by () below.
182 (i)
is rare in epic narrative; is more often corrupted into it
109

than vice versa, e.g. WD 756, Theogn. 750, 989. Phonetically

they were identical.
(ii) / * Flach
An absurd conjecture designed to obviate the quite unobjec
tionable hiatus, has no appropriateness to the sentence and
produces an unwelcome breach of Hermann's Bridge.
183 (i) Arundel 522
Banalization of the normal kind.
(ii) ante corr. : : Senensis I,IX,3
ante corr.
The prefixes -, -, -, are not infrequently confused,
would mean 'flew onto' (the earth), which is possible,
but it is better to have the connexion with 182 as given by -.
The form given by is incorrect, but shows the termination that
the scribe expected for a 3rd person plural,
(iii) Laur. 91 sup. 10 (corr. m 2)
Apparently just a misreading of iota as a narrow v.
184 (i) Vat. 1332 u: ' Casan. 356
The variation is limited to the r family, is not impos
sible, since the Doric scansion would have parallels in
Hesiod, but it is far more likely to be the regular banalization of
an unaugmented form. The reading of the Casanatensis should
not be described as a dittography: it is inherited from r,
with the further intrusion of a connecting particle, which is itself
a standard type of corruption,
(ii) V : Vat. 2185 (corr. m2)
One phonetic and one visual error.
(iii) ' om. V W X Casan. 356 Laur. 91 sup. 10 Vat. 1332
Arundel 522
The omission is common to a and r, though and u have
avoided it. It goes back to a scribe who was unprepared for a new
sentence in the middle of the line and who connected the parti
cipial phrase with what preceded. See below on 185 (i).
110

(iv) , * Et. G u d . :
Exeg.
T h e paraphrase in the a n o n y m o u s Exegesis may correspond to
' : can mean 'year' in Byzantine as in
modern Greek. Early epic usage allows either the singular or the
plural in such expressions, but not the dative. T h e reading of the
E t y m o l o g i c u m G u d i a n u m will therefore be a c o r r u p t i o n , of a
genitive plural presumably. Since the s u p p o r t for a genitive
singular is so uncertain, we shall naturally leave the plural in the
text.
185 (i) ' V X Casan. 356 Laur. 91 sup. 10 Vat. 1332
Arundel 522
T h e attachment of , to
led to the assumption of a new sentence here and the return of
the particle omitted in 184.
(ii) c (v in ras. Phillipps 11723): , ( ante corr.)
Casan. 356 Vat. 1332 U m S T r Q Laur. 31,32 m 2 Et. G u d . :
L : W
Laur. 91 sup. 10 Arundel 522 :
W a t 2185
T h e spelling with vv is apparently c o m m o n e r in manuscripts than
that with single nu. T h e transfer of the gemination to the rho may
have been an attempt to restore the metre after \ At any
rate it seems to g o back to the junction of a and r; individual
copyists corrected it (the scribe of X is seen in the act),
(iii) prius o m . Vat. 1948
Omission of inessential particle. T h e scribe saw
together.
(iv) U ante corr.
Anticipation of .
186 ' U ante corr.
Anticipation of .
187 (i) ' Q : Vat. 1948
0' can easily be misread as ', and ' immediately below may
have played a part.
Ill

() b ( superscr. L ; corr. M o s q . 469)

T h e elided verb was accidentally written in full. T h e corrector
of L, or a predecessor, misread the a p o s t r o p h e in ' as the
c o m p e n d i u m for at.
188 (i) Laur. 91 sup. 10 ante corr. Arundel 522
T h e w o r d may have been written in the exemplar.
(ii) O ' * r ( c o r r . U) T r , . L 1
F r o m ()' a b o v e . T h e change of subject calls for .
(iii) Vat. 1948
Probably a reminiscence of 108 ' and
113 .
189 (i) ' Laur. 91 sup. 10 A r u n d e l 5 2 2 : ' Casan.
356 Q c
T h e very unusual c o m b i n a t i o n was interpreted as the less
outlandish-looking by an easy visual error. T h e scribe w h o
w r o t e ' may have been thinking vaguely of 'horse'.
(ii) a r (corr. U)
Banalization.
(iii) X , W post c o r r . : Casan. 356:
V, U ante corr.
Following the c o r r u p t i o n to , s o m e o n e in the a family
tried to restore the metre with the impossible -. O t h e r
scribes assimilated - to m o r e familiar w o r d s : -
'famous', - 'closed' (quite unsuitable to the sea, of course,
b u t phonetically equivalent t o -).
(iv) . Casan. 356
T h e scribe thoughtlessly gives the w o r d the aspiration of the
m o r e familiar to him.
190 (i) S, Vat. 2185
T h a t we write but is pure c o n v e n t i o n .
H o w e v e r , the c o n v e n t i o n was established in antiquity, and these
scribes are perhaps t h i n k i n g of the m u c h c o m m o n e r modal
particle rather than of a legitimate alternative spelling of the a p o 112

copated form of . and are easily confused in minuscule,

and w o r d s ending in are exceptional.
(ii) Phillipps 11723
Assimilation of vowels perhaps influenced by t h o u g h t s of .
(iii) S post c o r r . : 2 : codd. ceteri: Fick
is clearly the paradosis, the t w o o t h e r variants being
copyists' emendations to repair the m e t r e ; b o t h S and the corrector
of X are p r o n e to e m e n d . As is a non-existent form, we have
the choice between and . Both forms occur with
in epic, t h o u g h only in later e p i c ; and there are
k n o w n cases of b o t h and being banalized to .
(Details in my c o m m e n t a r y . ) So no certain decision is possible.
(iv) : ex factum S : Phillipps 11723 ante corr.
I have not p u t ' S ante c o r r / , because I d o not suppose
that the scribe ever w r o t e . He probably corrected his error
after writing only . His eye may have slipped to below.
I have n o explanation for .
191 (i) o m . Laur. 91 sup. 10 (post rest, m 2 ) , Arundel 522
In an earlier copy in t w o columns written across the page, the
w o r d may have got transferred to the end of 190 (adjective and
n o u n t o g e t h e r : see below p. 138 on Catullus 61,219/20); that
w o u l d make its omission easy. T h e corrector of the Laurentianus
found it in the margin and, having n o u n d e r s t a n d i n g of metre,
inserted it next to its verb.
(ii) ' ' Senensis I , I X , 3
A n o t h e r intrusion of a connective at the b e g i n n i n g of a verse.
(iii) : in ras. A r u n d e l 522
Perhaps had been written by mistake, from in 190.
(iv) : in ras. Phillipps 11723: S
, i.e. , looks like a conjecture made because the end
of was illegible in the exemplar; or it may be a gloss, cf.
Hesych. * , * , , etc.
() : Senensis, Phillipps ante corr.
113

Assimilation to the next consonant, presumably.

192 Laur. 31,32
Phonetic dissimilation, or the influence of the present (see
above on 178 (iii)).
193 (i) T r Q cy . M o s q . 469
T h e rare ', guaranteed by the sense, was ousted by the
s o m e w h a t more familiar epic aorist of . Perhaps the man
responsible saw written in full, and t h o u g h t that it o u g h t
to be made into a dactyl.
(ii) (superscr. i) W : L
L's variant is merely orthographical, W's is a haplography.
(iii) ' : ex S ut id.
T h e scribe was thinking of , similar to in sound and
sense.
194 L e n n e p
There can be no objection to , and it is supported by
H y m n . Horn. 6,1 . . . Certainly a c o r r u p
tion of to would be explicable (as assimilation). But
is not the w o r d Hesiod uses for the genitals in this passage,
nor does he say that A p h r o d i t e developed in them, he says she
grew in the foam that formed r o u n d them. An unconvincing
conjecture.
195 (i) U 2 T r c (praeter Par. 2772; Laur. 31,32 m 2 )
A metrical e m e n d a t i o n ; Triclinius may have been its author.
But the p r o b l e m was already apparent to the scribe of S, and he
knew that the initial rho was sufficient to lengthen the syllable.
He indicated as m u c h by writing a second rho above it.
(ii) W : (accentu acuto postea eraso) Phillipps
11723
T h e scribe mistook the stem of the adjective for , and
made only a partial correction.
(iii) W : T r : Exeg.
114

T h e scribe of W was t h i n k i n g of <, , whence perhaps

also his --. T h e sigmatic element in the may also have sug
gested an aorist. T h e extraordinary , which is clearly
presupposed by the Exegesis (p. 382,23 Flach), must have origi
nated as a conjecture. O n e can only speculate on the reasons for it.
(iv) 8' om. Vat. 1948
196 (i) versum d a m n . Heyne
It interrupts the etymology ' - ' , and
forestalls that of . It looks to have been interpolated to
make the etymology more explicit with .
(ii) , codd., Et. Gen. et M a g n . : - Guyet, -
Werfer
T h e verse is ancient - it was apparently k n o w n to Clement and it is therefore likely that it was originally metrical. An accu
sative of fern, is indicated by echoes in O r p h . 11
and fr. 183,5. T h e c o r r u p t i o n , and the parallel one in 199, may be
explained by the influence of , twice nearby and , in other poets.
(iii) (sine ) T r
Triclinius omitted the particle in an attempt to mend the metre.
T h e result is a very r o u g h r h y t h m . He may have t h o u g h t more
capable of synizesis than ; or below may have affected
his pen.
197 (i) W ante corr.
A sort of h a p l o g r a p h y .
(ii) ' r : M o s q . 469 T r :
2833 Vrat. Rehd. 35 : ' Senensis
198 (i) Casan. 356

S : ' L Par.

Vat. 2185, ex vel contra

U (et marg. )
was probably the earlier reading in U, copied from u
(v/hich is a b r o t h e r of the Casanatcnsis and parent of the Vaticanus). This was a simple error of metathesis, and a conjecture
115

suggested by it and by the delightful atmosphere that Aphrodite

carries with her (cf. in 206).
(ii) r c
Banalization: is much the commoner epic form.
(iii) Arundel 522 : Vat. 2185
In the mental ear of a scribe w h o knew as a feminine
singular, kithtris naturally became . The other variant is
an echo of .
199 (i) u S Q, Laur. 31,32 m 2 , : V
r (praeter u) Mosq. 469 c (praeter Z) Et. Gen., . L 2 : -' Tr,
U post corr. : ( m 2 in ras.) : W Mosq.
469 post corr. et . : L
Par. 2833 Vrat. Rehd. 35 :
Schrevelius, - Werfer
- is incorrect; it may have been suggested by or
by other forms in - (-, -, etc.). The ending has suf
fered the same corruption as in 196 in part of the tradition, but is
correctly preserved by b. - is half and half.
(ii) 8' r S Q Mosq. 470 Phillipps 11723 Senensis Et. Gen. : '
V W b, ' in ras.: om. X post corr., U post corr., Tr Par. 2772
Laur. 31,32
8' is the more appropriate particle, especially after . 196 may
have helped to come in. The omission of the particle is a metrical
expedient at least in Tr U, where ' was written.
(iii) c (praeter Vat. 1948 ) : Vat. 2185 m 2 ,
Mosq. 469 ante corr., : ||||||||| : Vat. 1948
is a banalization of the unfamiliar . T o others the
lack of augment suggested (yento \yenito)\ they did not
unterstand the use of the optative in classical Greek. and Vat.
1948, by suppressing the verb altogether, contrive to make the
line scan despite the initial corruption.
(iv) schol. Horn., Et. Magn. :
(- V) b ( L 2 ) S Tr Q c (praeter Z),
ps.-Choricius, Et. Gen. : r (. U),
116

. L 2 : . Vat. S. Pietro C 152 :

, . [ . . . U
T h e context guarantees . is a reminiscence of 189,
and a further c o r r u p t i o n of it. is a deliberate
change to make the verse scan after ' .
It remains to choose between - and -. (For V's
see on 189 (iii); the fact that the same manuscript is
involved is typical. T h e suprascript in L may be meant as a
gloss.) is particularly appropriate to an island, and is
twice elsewhere applied to Cyprus, has an obvious
origin in 189 (whence also in r). Illud in hoc abiturum
erat, non hoc in illud.
200 (i) v e r s u m o m . u: m a r g . rest. U Vat. 2185
(ii) : 2 2 , U (marg.) v. 1., T r
A conjecture based on mis-scansion of .
(iii) S : T r c (praeter Laur. 31.32) : a r Vat. 2185 (marg.) L Par. 2833 Vrat. Rehd. 35 Q
Laur. 31.32: post c o r r . : , deinde [] ( post corr.) U (marg.) : superscr. M o s q .
469 : Et. G e n . cod. , : Bergk.
T h e appellation of A p h r o d i t e that Hesiod must be referring
to is (cf. 205 ). Because he derives it from
U r a n o s ' , the tradition has turned - into - or
(with the accent of the n o u n ) -. T h e suprascript in one of
the representatives of m will be due to conjecture rather than
tradition.
T h e simplification of the to caused metrical difficulties.
S may have restored the gemination by conjecture (cf. on 195 (i)).
S was available to Lascaris as he copied U, and it was no d o u b t the
source of the marginal restoration in U in its uncorrected form.
T h e corrector of fashioned the metrical b u t b a r b a r o u s hemistich
.
(iv) 2 : * Vat. 2185 : ' T r : ' Vat. 1948, '
: schol. H e p h a e s t .
117

OTTL, ' (of which ' and ' are corruptions), and 8*
represent further attempts to repair the metre. O n l y Triclinius
allows for a caesura; and his . . . is impossible in early Greek.
is an innocent variant for of a c o m m o n kind. It does not
suit the sense.
() Casan. 356 schol. Horn.
This must s o m e h o w have been generated by the correct , either by assimilation w h e n the latter stood in the text, or
by mistaking the reference of a marginal variant.
Apographa
Here are details of some a p o g r a p h a of manuscripts used above.
T h e Marcianus I X . 6 and the Salmanticensis 243 were b o t h
copied from before it was corrected. N o new errors were made
in lines 176-200.
C o n s t a n t i n o p o l i t a n u s 31 was copied from X after its correction
by X 2 . It has the following discrepancies from its m o d e l : 176
(corrected by the scribe); 181 (itacism); 187 ( omitted,
but restored by the scribe; 189 (re-banalization of the
conjectural - in X ) .
Scorialensis III 16 is a copy of U, with these divergences:
181 ; 185 ; 186 o m . ; 191 ; 193 (the
same c o r r u p t i o n as in T r Q c\ altered from ); 194
(influence of ', evi : evil?).
T h e Barocciani 60 and 109 were copied from the Scorialensis.
109 differs only in 181 (as if from instead of ).
60 has the banalization in 176, and makes in 194 into a
dative.
Mosquensis 462 is a copy of K. Its only difference in this passage
is in 178, w h e r e was at first written for / but then
corrected.
Marcianus 480 is a copy of Tr. T h e only new error is
for in 190. and are easily confused in some hands,
including Triclinius'.
118

Atheniensis 2965 and Glasguensis Hunter. U. 6.11 are copied

from Z. They agree in having in 198 (after , 192)
and in the banalizations 179 and or' for ' 200. In 178 had
with . m 2 ; the Athens copy keeps to ,
the Glasgow one has the altered to an by the original scribe.
In 199 the Athens copy after correction and the Glasgow one have
the which Z's reconstruction of the metre requires. The
Atheniensis has the further miscopyings 193 '; 195 ,
and ante corr.; 198 (cf. the note on the line, above);
199 ante corr.
Athous 3868 is also descended from Z, but through a Vatican
manuscript of which I have not got a collation. It diverges from as
follows: 178 om.; 179 misplaced at the end of 178; ;
181 (haplography); 183 ; 184 ; 185
; 186 ante corr. (from - endings in
185 and 187); 190 , ; 193 , post insertum
(as in W).
Two lessons can be drawn from these data. First, there is no
standard frequency of error. One scribe makes eight errors in a
passage where another makes none. Second, these copies some
times produce errors which coincide with ones known from other,
independent manuscripts: 193 Scorialensis; 178
Mosquensis 462; 198 Atheniensis; 193 Athous.
This is a serious reminder that the agreement of two manuscripts
in individual corruptions does not always prove affinity.

2. 'Hippocrates', de morbo sacro 1, 29-44

The first part of De Morbo Sacro is transmitted in two independent
manuscripts, (-) and M, of the tenth and eleventh centuries respec
tively. For details see H. Grensemann, Die hippokratische Schrift
"Uber die heilige Krankheit" (Berlin 1968). I have taken my in
formation about readings from Grensemann's edition.
119



,
5 ,
, ,
. , '
) .) ,
' .
, '

, 15 . fou .")
,
, ,
,
, ,
20 *
, ' .
, " .

, .
25 , .
' {
}. ,
30 ,
, ,
, ,
'
, ,
35 ,
20

, ' ,
.
1 (i) prius o m .
Omission seems a likelier error than insertion here,
(ii) scripsi : :
is the verb used of d r a w i n g d o w n the m o o n in line 8
and in other classical references to the a c c o m p l i s h m e n t (Aristoph.
N u b . 750, Plat. G o r g . 513a). In a H i p p o c r a t i c w o r k it may have
been written -, which w o u l d help to account for the variant
. T h e u n c o n t r a c t e d spelling - is c o m m o n in manu
scripts of Ionic p r o s e and verse, b u t historically incorrect.
2/3 L o b e c k , A g l a o p h a m u s i, 634 not. s:

C o m m o n sense c o m m e n d s the alteration. T h e r e are several
ways of explaining the assumed c o r r u p t i o n ; for instance,
might have been c o r r u p t e d to by assimilation and the
second then d r o p p e d out.
3
Anticipation of . Perhaps the Ionic - should
be written.
4 (i)
Banalization.
(ii) : (-)
T h e w o r d s are often confused, for graphical reasons, and here
the preceding xai's may also have exercised an influence, suits
the sense better.
5 (i) E r m e r i n s , : '

T h e impersonal goes ill with ,
suits it better and parallels
. After the initial c o r r u p t i o n had occurred, the scribe of
or a predecessor expanded the phrase to clarify its apparent
meaning.
121

(ii) N o t e that the c o m m a before might

instead be placed after it. But the phrase has m o r e point if it is
resumptive, s u m m i n g u p the lengthy protases.
6 W i l a m o w i t z ad E u r . Her. 1232 : ' ,
T h e preceding ^ are g o v e r n e d by , whereas is parallel to it and c a n n o t be attached by a third .
Assimilative c o r r u p t i o n .
7 (i) , corr. 2
(ii) (omissis - ) W i l a m o w i t z , G r . Lesebuch
I, 272 : ' :

T h e senseless in M must be nearer the original
reading than the version of , which clearly represents an at
t e m p t to restore sense, with eliminated and the clause c o n v e r t e d into a quite unsuitable final clause. If
is accepted as g e n u i n e (the only plausible origin for it), or
is u n a v o i d a b l e . T h e r e is no reason to suspect the clause, which is n o t in glossator's language.
8 (i) prius o m .
See on 1 (i).
(ii) c o d d .
See on 1 (ii).
9 (i)
A double itacistic error, characteristic of this manuscript. T h e
copyist may have felt that - was the sort of e n d i n g the ancients
were liable to use after 'if.
(ii) codd. : corr. W i l a m o w i t z
T h e w o r d division of the manuscripts is of course quite with
out authority. Y o u d o n o t say b u t
. T h e idiomatic is perfect.
10 (i) o m .
After the w o r d seemed r e d u n d a n t .
(ii) (-)
122

N o t impossible, b u t inelegant, and p r o b a b l y due to a scribe

consciously or unconsciously clarifying to himself the construction
of .
13 :
is merely O's careless o r t h o g r a p h y a g a i n ; 'este\ he said to
himself w i t h o u t t h i n k i n g what the w o r d s were, and it went d o w n
as if it were the future of . T h e article with seems neces
sary. It was easily lost after .
15 ir):
This sentence is so c o r r u p t that it is n o longer possible to see
even the general sense intended, and may b o t h g o
back to . If so, begins a new main clause: ' b u t
m o r e often they imitate these'. This may or may n o t be c o r r u p t ,
d e p e n d i n g on w h a t preceded. W h a t follows suggests that 'these'
are various animals. See also below on 16 (ii).
G r e n s e m a n n p r o p o s e s , exempli gratia,
, s u p p o s i n g the verb to have been c o r r u p t e d by the
following ,.. But it is hard to see h o w could have
b e c o m e () () .
15/16 - o m .
H o m o e o t e l e u t o n after .
16 (i) :
T h e true b e g i n n i n g of the sentence h a v i n g d r o p p e d out, the
scribe instinctively p u t in here.
(ii) W i l a m o w i t z : codd. : Erotianus ? : fort,
T h e parallel verbs are all in the singular (except for the variant
), of a single patient, and this one surely was t o o . T h e
transmitted plural will be due to the p r e c e d i n g ones. But it is
surprising that was n o t affected first: possibly
was a misplaced correction of it, t h o u g h w o u l d have
been the o b v i o u s conjecture. Erotian fr. 34 p . 109 N a c h m a n s o n
records as a Hippocratic w o r d , and there is a g o o d
123

chance that he had this passage in mind. If so, the plural reading
goes back at least to the first century.
or is to gnash the teeth, is to roar. T h e
writer p r o b a b l y used o n e or other form, t h o u g h there is some
evidence for the use of in the middle (Liddell and Scott
s. v. . at e n d ) , and G r e n s e m a n n reads . N o t e
that w o u l d easily b e c o m e . (before b e c o m i n g
) b e t w e e n and .
(iii)
natural assimilation to .
18 (i) (-): : corr. R e g e n b o g e n
An itacistic e r r o r in , a banalization in M .
(ii)
Perhaps right, b u t may well be an addition to clarify the con
struction. If it is right, its omission in m i g h t g o back to an
uncial copy ( - ) . Cf. 28/29 (ii).
19 (i)
A n o t h e r of this m a n u s c r i p t ' s mis-spellings; cf. above on 13.
(ii) scripsi : :
T h e conjecture accounts f o r k ' s impossible reading. 6 in may
be either a banalization of or a conjecture for .
(iii) :
may have w r o n g l y suggested a dative.
20 (i) G r e n s e m a n n : : o m .
It is easy to delete the nonsensical (so ), b u t we must ac
c o u n t for its existence. G r e n s e m a n n has recognized it for a cor
ruption of ', the n o m e n - s a c r u m abbreviation of .
(ii)
Unsuitable. Influence of ,
(iii)
Such variations are c o m m o n in prose a u t h o r s . T h e r e is n o t h i n g
to choose.
22 (i) scripsi : : (-)
124

A similar case to 19 (ii). There are five other places in De Morbo

Sacro where has , etc. and , (Grense
mann, p. 53). This can only be the result of a systematic process
on one side or the other. Here, has been assimilated to
the following noun.
(ii)
Cf. on 20 (iii).
24 : 'fortasse ' Grensemann
Erotian, in a group of entries in his Hippocratic glossary that
come from De Morbo Sacro, has the following:
.
f ( Welcker diagnostic conjecture),
"' ' " " ' '
" (frr. 334-5 Pearson).
The position of the entry suggests a location for the phrase
between 1.10 and 1.40 , and the passage about
gods certainly offers a promising field in which to seek it. Since
the causes a in Sophocles, and elsewhere is
equated with Hecate, Grensemann proposes (first in Hermes 93,
1965, 490) to place Erotian's here, and to assume
that is a gloss.
Two objections may be made. One is that a masculine
was unlikely to be glossed by the name of a goddess.
in 19 was much likelier to be glossed . One might
suppose that was originally intended to refer to that,
and accidentally displaced instead; but the second
objection is more serious, that if Erotian's entry came from here
it ought to be in the genitive or in the nominative. We should try
to place it rather where there is room for an accusative. Since an
seems essentially to be one who is supplicated in
adversity (, Hesych. and Et. Magn. s. v. ;
sch. . Rhod. 1.1141), a suitable
place would be in line 30 where we read , which could
125

be a banalization. It is true that this is just after (28),

which follows in E r o t i a n , but it is not the case that
strict order is preserved, as a glance at N a c h m a n s o n ' s edition
shows.
26 o m .
T h e w o r d is clearly genuine. A case of simple omission, pos
sibly helped by the fact that the next w o r d begins with the same
letter.
27/28

It is difficult to say anything for or against the articles with
. But the Ionic - seems to have impressed
the scribe so m u c h that he involuntarily w r o t e - for b o t h
before and after, is impossible and unlikely. F o r
the omission of c o m p a r e 1 (i), 8 (i).
28
and b o t h existed in fifth-century Ionic,
but the former is slightly the difficilior lectio, cf. Hesych. * , . , , ,
.
28/29 () del. R e g e n b o g e n , Symbola
Hippocratea p. 3
R e g e n b o g e n points o u t that E r o t i a n ' s explanation of
includes the alternative
. T h e phrase in o u r text is clearly an intrusive version of the
same interpretation. Such glosses are often introduced by ('in
other w o r d s ' ) , and here the phrase b e g i n n i n g with seemed to
belong in the sentence,
(ii) :
An unintelligent uncial m i s r e a d i n g ; cf. on 18 (ii).
2 9 (i)
(ii)
Difficult decisions. O n the one hand one may argue that
is s u p p o r t e d by D e Aere Aquis Locis 5,1
126

 ( the same writer as D e M o r b o Sacro); H e r o d o t u s 6,86a,l

, 109,6; 8,102,2; 9,104; and that is a mere
haplography. O n the o t h e r hand H e r o d o t u s can speak of a
(7,10,1), and while this is a less similar phrase,
it makes it hard to question the possibility of :
and could b o t h easily be u n d e r s t o o d as lectiones
faciliores. N o r can or confidently
be excluded.
(iii) o m .
T h e implies a verb linked m o r e closely with than
can be- is ideal. H o m o e o t e l e u t o n explains the omission.
30 : fort,
See above on 24.
31 J o n e s
is usually the rite and the concrete object of
it. T h e neuter . . . also favours J o n e s ' s conjecture, and
it is palaeographically easy, since stems in -- are often ab
breviated in minuscule, . O n the other hand H e r o d o t u s
speaks of ' m a k i n g A t h a m a s a ' (7,197,3), and Plut. de
curiositate 6 p. 5 1 8 b has .
32 (-)
T h e sense is n o t affected, - might in theory be an anticipa
tion from , but it is m o r e reasonable to regard it as the
original reading and as a simplification.
33 (i) (.
T h e p r o d u c t of uncertainty a b o u t classical syntax,
(ii) : .
T h e sense d e m a n d s a , and the reading of evidently goes
back to . Between - and - there is n o t much to
choose, but - (which perhaps carries a stronger implication of
i n v o l v e m e n t in what is t r o d d e n on) appears in connexion with
defilement in Heraclitus fr. 86 Marc, and Soph. O C 400.
34 (i) 6 (-): : possis :
127

The articled matches the preceding . This corres

pondence may be due to the author; or on the other hand to a
scribe, if the author meant 'render them to the god concerned,
if a god is responsible', . . . seems to be unparalleled;
would be paralleled by variants in Herodotus 7,10,1 and Plato
Parm. 135 b, see Denniston, Greek Particles 247; and it would
account for the presence of in M.
(ii) (-)

The writer always uses in expressions of opinion except

when he has enclitic forms of the pronoun (2,1 )
or when there is a connective to be accommodated (1,10
,). The examples may be found from Grensemann's
index. 0's reading is a banalization paralleled at 17,2 ( :
C, a 15th-century manuscript which is of use later in the work).
35
There is no such word. The influence of one or more of the
neighbouring ' is responsible.
36 (i)
A merely orthographical variant. The author probably said
(or ) in reading out his work, in writing he may have put
either.
(ii) ante habet M
The sentence was too long for the scribe. He did not realize
that etc. was still governed by , and felt the need
to supply another verb, which he did without much understanding
of the sense.

3. Aesop, fab. 157 Perry; Xenophon, Memorabilia

1, 3, 8^9
These short items are included here not so much to illustrate the
textual critic's way of working as to show how freely prose texts
can be altered at the hands of copyists. In the Aesopic collections,
128

which were valued for their substance and n o t for any stylistic
reason, variations of w o r d i n g from manuscript to manuscript are
the rule, t h o u g h often not unconnected with c o r r u p t i o n by mis
reading. T h e fable of the fox and the goat may serve as an example.
It is transmitted in the following m a n u s c r i p t s 1 ) :
(sigla and g r o u p i n g s )
(century)

G Pa Pc.
xii xiv

Ca M b .
Pb
Pd. Ma M o
xiii xv
xiii/xiv xv xv xv

,
, ,
, '
. "'
5 , ".
'
.
1 G Pa Pc
T h e oldest g r o u p of manuscripts on the whole offer the better
text, b u t this cave is not very appropriate.
2 (i) G : Pa P c : . .
M b P b Pd M a M o , . Ca
T h e placing of next to in the later manu
scripts represents a simplification of w o r d order.
(ii) Ca M b
3 ' M b et omisso
Ca P d : ' P b :
' M a : '

W h e n the readings are presented in this o r d e r (which happens
]

) I have used Perry's sigla, and constructed my apparatus from his edition and
Hausrath's clumsier one.
129

to be chronological), it is not difficult to see their genetic relation

ship, was c o r r u p t e d , probably t h r o u g h
(the scribe expecting another comparative), into .
Seeing the absurdity of 'violent' grass, a n o t h e r copyist emended
to . In a n o t h e r branch of the tradition
was c o r r u p t e d into ( M a ) ; M o makes
sense of this by o m i t t i n g and substituting another adjective.
complete stemma lectionum will look like t h i s :
^ ' ( ; , | ^ . G P a P c
7
" () |[] .' |[|

CaMbPd

Pb

'

' -:

4 G Pa: P c : cett.
5 (i) Pc
(ii) : ' P c : Ca M b
6 G Pa: P c : cett.
comes from the following verb and is unlikely in con
junction with it.
7 (i) Pc
(ii) Ca M b : G Pa Pd Ma M o , () P b ,
Pc
T h e sense guarantees . Predictably it was mistaken for
the much c o m m o n e r w o r d . Pb's for
130

 is a further misreading, probably influenced by .

O n the other hand, while / is a c o m m o n graphic confusion
(in early minuscule, particularly), Pc's reading looks like a con
jecture, meaning ' i n v e n t i v e ' , which is not a p p r o p r i a t e to the ending
of the fable or to .
(iiiJ om. Ca
Haplography?
A manuscript tradition that does not display such variations
as these offers less of a challenge to the critic, and he may be
beguiled into excessive trust in its fidelity so long as sense and
style seem to run smoothly. T a k e this passage from X e n o p h o n ,
Memorabilia, 1,3,8-9:
*
.

, * " "
" , avOpco
;"
Why should anyone d o u b t that this text given by the medieval
manuscripts, and as far as also by Stobaeus 3,17,43, is
a faithful r e p r o d u c t i o n of what X e n o p h o n w r o t e ? Except that we
happen to possess a fragment of a copy made a b o u t one century,
instead of eight or sixteen centuries, after X e n o p h o n , and it gives
the text t h u s :
[ ] [ ]
[ ] [ , ]
[
] . [ ]
[ ] [ ' ] ,
[ ' ] [ ]
[- . . . .
(. Siegmann, Literarische griechische Texte der Heidelberger
P a p y r u s s a m m l u n g , 1956, no. 206.) T h e version of the papyrus is
not necessarily authentic in every detail. T h e phrases that it omits,
131

 and , look genuine.

But so does the extra clause after ; it would be arbitrary
to dismiss the Ptolemaic text as aberrant and cling throughout to
the Byzantine one, which must after all have Ptolemaic ancestors.
However one judges the differences, the unity of the tradition
has been proved illusory. How many other variants may not have
been current in ancient copies of these lines?

4. Catullus 61, 189-228

The extant manuscripts, about 110 in number, all descend from
a single copy which was written perhaps in the ninth century,
discovered at Verona in the late 13th contury, and subsequently
lost. Only three of the whole swarm are primary witnesses, G R,
all of the 14th century: the rest are derived from them. As far as
they are concerned, it is a closed recension, the stemma being:

190

at marite, ita me iuuent

caelites, nihilo minus
pulcher es, neque te Venus
neglegit. sed abit dies:
perge, ne remorare.

189 Ad maritum tamen iuuenem G R (sc. V ) : corr. Scaliger

The transmitted reading is unmetrical and nonsensical, caelites in
190 requires a construction; and what follows shows that the bride
groom is being addressed, as in the preceding stanza. That leads
us to marite (as in 184), and it follows that ad conceals at (often
written ad in manuscripts and inscriptions in spite of the gram
marians' rule given by Quintilian 1,7,5 and others; cf. below on
132

225). In the remainder, tamen iuuenem, Scaliger recognized a formula

like 66,18 ita me diui .. iuuerint, 97,1, ita me di ament. It is hard to
say exactly how the corruption came about, ad might have gen
erated the accusatives,TnArvco-being read as . Or tuue-nr
might first have been read as iuuen? and the corruption spread
back from it.
190 (i) Celites
(ii) nichil ominus : nichoilo minus G
The words nihil and mihi are commonly written nichil and michi
in manuscripts. The scribe of G anticipated the ending and wrote
nicho, then realized his error and corrected it.
191 (i) Pulcre res V: corr. 'alii' ap. Robortellum
Again nonsense, though arranged in Latin words. It is clear from
the context that the groom is being complimented in the conven
tional way. The Catullus manuscripts usually spell pulcer without
an h; this corruption may go back to either spelling. Pulcher was
the usual pronunciation in Catullus' time (Cic. Orator 160), but
it could still be written pulcer, and the manuscript evidence may
indicate that he wrote it so, if it does not reflect a deliberate
archaizing policy by scribes of the imperial period.
(ii) nee V: corr. rec.
192 abut V : corr. rec.
The same phrase (sed) abit dies has been used in lines 90, 105 and
112, and V has the same corruption each time. So has X in 63,38,
the only other place in Catullus where abit occurs, while in 63,30
one copy has abiit for adit.
193 rememorare X
Perhaps a mechanical dittography like 'renonown' (above, p. 24),
but more probably the thought of the word memorare, or an actual
variant, played a part; cf. Seneca, Quaest. nat. 2,55,4 remorari
: remorare HPG : memorare F: rememorare T. Note that the
impossible form was faithfully copied in both G and R.
133

195

non diu remoratus es:

iam uenis. bona te Venus
iuuerit, quoniam palam
quod cupis cupis et bonum
non abscondis amorem.

194-8 ante 189 habuit V : tramp. Scaliger

The stanza goes well enough after 184-8 iam licet uenias etc., but
189-93 makes no sense afterwards: nihilo minus has no reference,
and ne remorare does not come well after non diu remoratus es.
Scaliger's transposition restores coherence between each of these
three stanzas and the one following it, and is unquestionably right.
194 remorata X : remota corr. Calphurnius {ed. 1481)
V must have had -ta; and it will have had remorata, since if it had
had remota, that would also have appeared in X. remota is a second
ary corruption in O, a quasi-haplography.
The true reading is obvious, but to scribes with little under
standing of metre and a limited field of vision it was not quite
so obvious when the line immediately followed 188. It was the
sentence about the bride, no doubt, that generated the feminine
ending.
195 Iam uenus recc. duo
A nice example of assimilation.
196 Iuuerit rec. : Inuenerit V : Inuenit rec.
Of the two humanist emendations of V's impossible reading,
iuuerit is far superior to inuenit from the point of view of the sense,
making an expression like those mentioned on 189. Once iu- was
read as the much commoner syllable in- (a very easy visual error),
the unrecognizable remainder -uerit might naturally be made into
-uenerit, especially with uenis directly above. Or inuenit may have
been written and then conflated with a correction.
197 cupis capis G, R m. rec.
This reading, which gives good sense, was easy to think of, and
it would be rash to assume that the agreement of G with a corrector
of R implies that it was a variant in X. cupis cupis must be treated
134

as the paradosis, and it is unexceptionable. The other must be

treated as an emendation, and as such it is a worthwhile one.
Corruption by assimilation would have been very easy.
198 abscondas V : corr. recc.
Grammar and metre impose the indicative. I cannot account for
the corruption, unless the a comes from the variant to the eleventh
letter of 197.
ille pulueris ericei
200 siderumque micantium
subducat numerum prius
qui nostri numerare uolunt
multa milia ludere.
The reader to whom the emendations so far discussed have appeared
simple now has the opportunity to try his own hand, unless he
remembers how this stanza reads in a modern text. I have given
it as it is transmitted, and postpone my discussion of it to the end
of the present section.
205

ludite ut lubet, et breui

liberos date, non decet
tarn uetus sine liberis
nomen esse, sed indidem
semper ingenerari.

204 Et ludite et lubet V : corr. Calphurnius (ut), Parthenius

ut and et are often confused. In the present case assimilation
comes into play too. The superfluous et before ludite may come
from an instinctive tendency towards symmetry that turned et
et C into et A et et C.
208 ingenerati
visual error.
2io

Torquatus uolo paruulus

matris e gremio suae
porrigens teneras manus
dulce rideat ad patrem
semihiante labello.
135

209 Torcutus O
This is the only place in the poem where the name appears, but
there can be no doubt that Torquatus is the correct form of it.
The copyist of probably saw the a as a u, lorquutus.
210 et
Another common confusion, due to abbreviation of et.
213 Sed mihi ante V (michi G) : corr. Scaliger
The nonsensical transmitted reading was easily cured once the
word division was disregarded. The unusual long word had been
seen as three much commoner words. G then modified the spell
ing of mihiy cf on 190 (ii).
215

sit suo similis patri

Manlio, et facile inscieis
noscitetur ab omnibus,
et pudicitiam suae
matris indicet ore.

214 simili rec. unus m. pr.

A typical assimilation-corruption.
215 (i) Maulio : Mallio rec.
Maulius is not a Roman name, and clearly derives from a mis
reading of Manlio, which therefore stood in V. The name is also
given in line 16, where V had Mallio. Mallius is a genuine name,
often confused with Manlius, Manilius, etc. Corruption in either
direction was easy, since Manlius was the better-known name,
while Italian pronunciation tended towards Mallius. The former
is made probable here by Torquatus in 209, for the Manlii Torquati
are a well-attested family. The man who wrote Mallio no doubt
remembered 16.
(ii) ut Bergk
A quite legitimate suggestion (cf. on 204).
(iii) facie Burmann
Another conjecture based not on any objection to the transmitted
reading but on the discerning of its possible origin from some136

thing the. facie nosatare is good Latin (Livy 22,6,3), but when we
have ore in the next clause it is labouring the point too much to
have facie as well; it adds nothing to the sense, whereas facile does
add a little.
(iv) insciis R m. rec, -eis Lachmann : insciens V: vix insolens
The infant's own ignorance is irrelevant, and insciens cannot be
right. insci(e)is adds point to 216: 'may he be recognized as his
father's son even by those who do not know'. The spelling -eis
was still in use in the late Republic, and there are distinct signs
of it in the manuscripts of Catullus and Lucretius. This is one
place where the assumption of its use helps to explain a postulated
corruption.
One might also think of insolens, 'though he is unfamiliar to
them', but I am not sure whether the adjective can be used in quite
that sense.
216 omnibus cum insciis commutat Dawes : obuieis Pleitner
217 (i) set lland\ sic Hermann
All these conjectures are attempts to eliminate the anomaly of
the short syllable at the end of 216. Synapheia prevails otherwise
throughout the poem, as in Catullus' other glyconic composition,
34: the last syllable of the line is genuinely long except at the end
of the stanza, and words may be elided at line-end or even broken
between lines. The material is sufficient to establish the rule, so
that the problem is a real one. Dawes's transposition is not very
felicitous; inscieis is better in the predicative position, the desired
emphasis being ab omnibus, etiam inscieis, not ab inscieis, et quidem
omnibus. Pleitner's obuieis, as an adjective in substantival use, could
hardly be qualified by another adjective such as inscieis. There
seems more hope in attacking et in 217, a word often involved in
corruption. Hand's set (i.e. sed) is contrary to the sense. Hermann's
sic is more suitable, only the clause is so nearly equivalent to the
two preceding that one hankers after a closer connexion. No convincing solution seems yet to have been found, and as the scope
137

for emendation appears so limited, it may be that having stated

the anomaly we should with due reserve accept it.
217 (ii) pudiciciam
ci and //, being both pronounced tsi in the Middle Ages, are liable
to be interchanged. Hence such spellings as conditio for conditio,
nuncius for nuntius.
(iii) suam V : corr. Calphurnius: suo R corr.
The required sense is clear. It could be given by suo, but the
word-order is then involved, and the other correction is supported
by 58 a gremio suae | matris, 210 matris .. suae, 214 suo .. patri. The
corruption is a commonplace assimilation of endings.
218 iudicet
The sense guarantees indicet. The easiest of graphical errors.
talis illius a bona
220 matre laus genus approbet
qualis unica ab optima
matre Telemacho manet
fama Penelopeo.
219/220 bona matre/Laus V
similar misdivision occurred at 58, a gremio suae matris/'instead
of suaejmatris. A scribe's eye ran on from the adjective to the noun
that it heralded.
221 ab om.
222 thelamacho : theleamaco X
Uncertainty over aspiration (t/th, c/ch) manifests itself particu
larly in Greek names. The variants in the second syllable point to
V's having had thela- with a suprascript e over the a. The a was an
anticipation of the third vowel.
223 penelopeo O, penolopeo X : Penelopaeo Housman, J. Phil. 33,
1914,73
penolop- is the same kind of mistake as thefom&cho.
A name in - normally gives an adjective in -ouoc, and Housman
observes that the transmitted Penelopeus here and in Ovid Tr. 5,
138

14,36 may stand for -paeus. However, the question is not a simple
one. We do not know that the adjective was ever used in Greek.
If Catullus first coined it, the question is what analogies were
uppermost in his consciousness. It is not a question that we are
well placed to answer; but it seems possible that the poetic form
of the name itself, , might have made Penelopeus sound
right.
225

claudite ostia uirgines:

lusimus satis, at bonei
coniuges bene uiuite et
munere assiduo ualentem
exercete iuuentam.

224 hostia V
225 (i) ad V: exp/. recc.
See on 189.
(ii) bonlei : bolnei X (al. bonei R,G corr.)
V perhaps had bonei with a suprascript / that looked like an / to
both copyists.
226 bone uite V : corr. R m. rec. : bone uitas, bonae uitae {partim
sine et) recc.
No part of uita will fit here, and bene uiuite, giving a verb in para
taxis with exercete, is a certain correction. Haplography produced
uite, while bene was affected by bonei and became an adjective
qualifying uit{a)e. Attempts were then made to accommodate the
noun in the construction by making it uitas et .. iuuentam, or bonae
uitae munere.
227 assidue V : corr. recc.
munere needs an adjective to give it meaning, -e may have been a
mechanical assimilation to preceding endings.
228 Exercere
An easy corruption in purely visual terms, but munere was probably
a contributory factor. As a matter of method, one ought to con
sider the theoretical possibibility that exercere stood in V (-ete in
139

X being a natural correction) and that the corruption really lay

elsewhere. Once the question is asked, the idea quickly suggests
itself that ualentem, or ualete, might conceal ualete. But while
ualentem in this context might well become ualete, corruption in
the opposite direction is difficult to imagine. An adjective with
iuuentam is also supported to some extent by Phaedrus app. 12,4
et exercehat jeruidam adulescentiam; Calpurnius Siculus 5,11 pnauam
potes exercere iuuentam.
Postponed discussion of 199-203:
The stanza does not make sense as it stands. The two last lines
are unmetrical as well as syntactically incoherent, while ericei is a
vox nihili. We must begin by ascertaining what sense was intended.
Evidently something like Met him first reckon up the number of
the sands and stars, who would count the myriad games of love'.
Compare 7,3-12 (with 5,10 for milia multa). We proceed to detail.
199 ericei: an adjective qualifying pulueris seems to be required.
aridi (Broukhusius) is palaeographically not hard, but very banal.
Renaissance scholars more laudably sensed that a geographical
term was wanted, and developed ericei through erit{h)ei into
Fzrythrei. But then metre had to be mended by reducing pulueris
to pu/uis, an impossible form of the genitive; in any case -ei cannot
stand for -aei. These efforts were superseded when Heinsius performed the simple equation, ericei = AURIC I = AFRICI, confirmed absolutely by the parallel passage with its Libyssae harenae
(7,3). Lachmann wrote Africei, and this spelling seems to have
been used in at least one ancient text of Catullus. It may be a
scribal archaism, or it may go back to the poet (cf. on 215 (iv),
225 (ii)): it had no etymological justification in the genitive singular as it had in the nominative and dative/ablative plural, and
Lucilius condemned it, but the distinction in pronunciation disappeared and in Catullus' time original / was sometimes being
written ei.
(200 micancium ( ) : see on 217 (ii).)
202-3: uolmit is unmetrical, therefore corrupt; but it is the most
140

suitable verb. The only part of it that would scan is uolt. Since
ille .. subducat .. qui .. calls for a singular, that is clearly right, it
was read as uoit, which occurs as an abbreviation of uolunt.
ludere is also unmetrical. noun in the genitive appears to be
required: ludi, or ludei. The ending was assimilated to numerare.
One word remains incongruous: nostri. The adjacent stanzas
show that Catullus means the love-play of the bridal pair: uestri
or uostri. If the word was written in full, the latter form was the
more easily misread; but it may have been represented by the
nomen-sacrum abbreviation lift, which often causes confusion.
How did you score? If you failed to see that the words ericei,
nostri, uolunt and ludere were corrupt, I see little hope for you as a
textual critic. If by your own unaided wit you restored uestri (or
uostri), uolt (or uult) and ludi (or ludei), you are competent. If you
restored Africei, you are brilliant.

5. Ovid, Amores 3, 15
Whereas the tradition of Catullus issues from a single source after
the late 13th century, that of Ovid ramifies throughout the
Middle Ages, and gives us better opportunities of observing what
liberties scribes can take with a text. As one might expect in the
circumstances, it is an open tradition. The four oldest and best
manuscripts of the Amores are R of the 9th century, of the 9th
or 10th, and S of the 11th. For 3,15, however, the last poem of
the collection, neither R nor S is available; and ends at line 8.
For the rest we are dependent upon and upon manuscripts of
later date (the oldest being of the 12th or 13th century), known
collectively as and going back to a source that is at least some
times independent from that of R S Y. I have used the sigla of
Kenney's Oxford edition with the addition of Munari's J and the
newly-respected (on which see Munari, II Codice Hamilton 471
di Ovidio, Rome 1965). The sign - means 'some of the <o manu
scripts'.
141

10

15

20

Quaere nouum uatem, tenerorum mater Amorum:

raditur haec elegis ultima meta meis;
quos ego composui Paeligni ruris alumnus,
nee me deliciae dedecuere meae.
{si quid id est, usque a proauis uetus ordinis heres,
non modo militiae turbine factus eques.}
Mantua Vergilio gaudet, Verona Catullo:
Paelignae dicar gloria gentis ego,
quam sua libertas ad honesta coegerat arma
cum timuit socias anxiaRoma manus;
atque aliquis spectans hospes Sulmonis aquosi
moenia, quae campi iugera pauca tenent,
"quae tantum" dicet "potuistis ferre poetam,
quantulacumque estis, uos ego magna uoco."
culte puer, puerique parens Amathusia culti,
aurea de campo uellite signa meo;
corniger increpuit thyrso grauiore Lyaeus,
pulsanda est magnis area maior equis.
imbelles elegi, genialis Musa, ualete,
post mea mansurum fata superstes opus.

2 (i) traditur ( macula obscuratur) : corr. Heinsius

Traditur meta does not make sense. But meta is sound; Ovid uses
the same image in Ars Amat. 1,40; 2,426. The required sense for
the verb is that expressed in those passages by premere and terere^
and in Am. 3,2,12 by stringere. Heinsius saw what traditur pointed
to. A scribe had substituted a commoner word and one more
readily connected with elegi. The occurrence of tradita at the be
ginning of a pentameter sixty lines earlier may have contributed.
(ii) haec ( obscuratur) : hie Pa
Hie may be only a conjecture in the two 13th-century manuscripts
from which it is quoted; if it were the only reading attested no
one would quarrel with it, but haec is hardly to be explained away
as a mechanical (even less as a deliberate) assimilation,
(iii) meta P Y -": nota H: c(h)arta -: cura Pb
142

After the corruption of raditur, meta became puzzling, and charta,

cura are deliberate substitutions by copyists; nota, however, looks
like a visual corruption of meta.
5-6 (^Trist. 4,10J-g) post 10 habent E a O b : susp. J. Schroder, lib.
emend. 205
The virtual repetition of a couplet in another poem is not in
itself a sign of interpolation in one or other place, But when there
is added to this the fact that it is not given by all manuscripts in
the same position, and the fact that in the Tristia it sits well in an
extended autobiographical passage, while here it interrupts con
centration on Ovid's place of origin, and breaks the connexion
between 4 and 7-8, then it certainly becomes worth considering
the hypothesis that the lines are not genuine here but were added,
at first in the margin, by someone who saw fit to append a detail
from another autobiographical poem, much as another couplet
from the Tristia (3,3,73-4) is added at the end of the elegy in one
manuscript (X). The question is complicated by something else:
6 modo militiae turbine to : ego fortunae munere E a Ob modo
fortunae munere Ov. Trist.
The same manuscripts that have the couplet in a quite unsuitable
place offer a major variant that agrees with the Tristia version
(except for ego). The combination strongly suggests that in these
manuscripts at any rate the couplet comes from a marginal quota
tion of the Tristia. But this need not mean that it does not be
long in Am. 3,15, after verse 4: its presence there might itself
prompt the marginal quotation of the parallel, and the reception
of the latter into the text after 10 might be accompanied by the
deletion of the original couplet after 4. So only the argument
from inner coherence can be used against it; how much weight
that carries, each reader will decide for himself, but it is enough
for me.
Militiae turbine is a more precise and forceful phrase than
fortunae munere. One can imagine copyists replacing it by the
other, but not the contrary. So it is possible that Ovid wrote
143

militiae turbine in the Tristia. T h e assumption is certainly helpful

if one regards the lines as an interpolation in the Amores.
8 gentis gloria dicar T h e transposition is of the c o m m o n type that brings related
w o r d s t o g e t h e r : Paelignae gentis, dicar ego. T h e metre is preserved;
scribes could cope with elegiacs as they could n o t (till the renais
sance) with Catullus* glyconics.
9 q u a m E a Q : quern T h e antecedent is gentis, but scribes mistakenly took it to be
the more emphatic ego.
\2 parua Perhaps influenced by 14. Pauca is s u p p o r t e d by Fasti 3,192.
13 dicit, dicat Such variations are c o m m o n . T h e subjunctive may have been
introduced in the belief that dicar in 8 was a subjunctive. It is
better taken as a future (cf. H o r . Carm. 2,20,3 ff.; 3,30,6 ff.), and
that confirms dicet here.
15 amathusia culti corr., -: amat uisia culti (c ex u facto) Y :
amathontia culta H : amat hostia cultum F, -us X: mihi t e m p o r e
l o n g o cett.
Venus is clearly being addressed, and Amathusia culti, even if it
is an emendation, is indubitably right. T h e Greek name, as so
often, baffled the copyists. FX have d i s m e m b e r e d it to form
Latin w o r d s , which h o w e v e r make nonsense. Someone else,
desiring sense, substituted mihi tempore longo. It is not very g o o d
sense, and if this had been the only reading preserved, critics
would certainly have queried it. O t h e r s w o u l d have defended it
as perfectly O v i d i a n , pointing irrelevantly to other examples of
tempore longo in O v i d . But w h o could have guessed the true
reading?
18 est magnis : Haemoniis Merkel. F^mathiis Martinon
T h e r e has been felt to be s o m e t h i n g flat a b o u t magnis, especially
with area maior. MerkeFs conjecture is based on P r o p . 2,10,1-2,
144

sed tempus lustrare aliis Helicona choreis | et campum Haemonio iam

dare tempus equo. The assumed corruption would not be difficult:
a copyist expecting est after pulsanda might well see it (abbreviated)
in the e of emoniis, and the mysterious residue might well be made
into magnis. Martinon's alternative has similar qualities, without
the advantage of the Propertian parallel. Haemoniis may be right;
at least it deserves a place in the apparatus. But magnis also has
point, and it is precisely the juxtaposition with maior that gives
it point. 'Big horses must have a bigger space to gallop in'; in
other words, have great powers, I must exercise them on some
thing more ambitious than elegy'.
19 Musa : turba
The use of turba with an adjective as an appositional phrase
is a favourite mannerism of Ovid's, as e.g. in Fasti 4,764 et ualeant
uigiles, prouida turba, canes. If it were the only reading transmitted
here, it would not be suspect. Given the much better supported
variant Musa, one may hesitate. It is not obvious why either
should have been replaced by the other.
post 20 sequitur sine intervallo in X Hie ego qui iaceo tenerorum
lusor amorum | Ingenio peril Naso poeta meo (Trist. 3,3,73 sq.),
in J Hie tua iam Naso deponit castra Cupido, | Hie dat militiis
ultima signa tuis.
Clearly neither couplet belongs here. They were added to
embellish the end of the book.

6. Apuleius (?), de Platone 2, 20

The somewhat corrupt tradition of this work rests on six manu
scripts, BMVPLF. is the oldest (11th century), and quite often
the only copy to preserve a true reading or a trace of it. The other
five are dated to about the 12th century. L F derive from a
hyparchetype , which sometimes has a better text than V; in
certain cases this can be ascribed to emendation, but in others a
145

better tradition must be assumed. F is particularly given to

emendation itself, and the cases where it alone gives the right
reading are so to be accounted for. This behaviour by F often
leaves PL on their own; there are indications, however, that
their shared readings do not always go back to 8, but that there
was another intermediary. The two remaining copies MV also
derive from one exemplar, whose text generally resembled that
of B.
These facts can be accommodated in either of two stemmas:

Since 8 generally offers a more intelligible text than MV, it is

more likely than MV to be the product of selection from alternative sources, which is a point in favour of the first stemma.
perfecte sapientem esse non posse dicit Plato nisi ceteris ingenio
praestet, fartibus et prudentiae partibusf absolutus atque fenim
iam turn a pueris inbutus, factis congruentibus et dictis adsuetus,
purgata et efficata animi uoluptate, electis ex animo <^de)hinc
5 abstinentia atque patientia omnibusque doctrinis ex rerum scientia
eloquentiaque uenientibus. eum qui per haec profectus fidenti et
securo gradu uirtutis uia graderetur, adeptum solidam uiuendi
rationem frepente fieri perfectum: hunc repente praeteriti futurique aeui ultimas partes adtingere et esse quodammodo intem10 poralem. turn post hoc, uitiis exclusis insertisque et inmissis
(uirtutibus), omnia quae ad beatam uitam ferunt non ex aliis
pendere nee ab aliis deferri sibi posse sed in sua manu esse sapiens
recte putat. quare nee in secundis rebus effertur nee contrahitur
in aduersis, cum se ornamentis suis ita instructum sciat ut ab his
146

is nulla ui segregetur. hunc talem non solum inferre sed ne referre

quidem oportet iniuriam; non enim earn contumeliam putat quam
inprobus faciat, sed earn non putat quam patientia firmiter toleret,
qua^ndo^quidem naturae lege in animo eius sculptum sit quod
nihil horum possit nocere sapienti quae opinantur ceteri mala esse.
2 (i) praestet : praestat BMV: 'fort, praestet, <et>' Thomas
Grammar calls for the subjunctive; 8 may have restored it by
conjecture. The addition of et is of course possible, but there is
nothing particular to be said in its favour.
(n) artibus et prudentiae partibus: malim prudentiae artibus vel
artibus et prudentia a parentibus
prudentiae partibus is an odd phrase, and the paronomasia with
artibus uncharacteristic of the style, artes is corrupted to partes in
2,21 and perhaps 25, so that one possibility is that prudentiae artibus
is the true reading (cf. 2,14 bonis artibus, 21, 22), artibus being a
correction of partibus that entered the text and generated a con
necting et. Or artibus may stand unqualified as in 2,26 easdem
puerorum nutricationes, easdem uult esse artium disciplinas, followed by
et prudentia a par<en>tibus, which would balance iam turn a pueris,
though it would clash with the account given later of the com
munal upbringing of children (2,25-6, from the Republic).
(iii) enim iam: iis iam Scaliger: etiam Sinko: eximia <(disciplina)
(et statim pro turn) Novak: malim enixim iam
enim is impossible; there is no reason to suspect iam. For the
sense one might expect something meaning 'continuously' from
childhood, or 'vigorously'. If enim is to be changed into an adverb,
it is reasonable to look first for a suitable adverb in -im, and
enixim suits both sense and palaeography. But it cannot be regarded
as more than a guess.
3 (i) puero F
Classical Latin uses a puero when the subject is singular, a pueris
when it is plural, but the plural here seems more likely to be a
genuine variation of the idiom than the result of corruption.
F's reading will be a conscious or unconscious correction.
147

(ii) et dictis: edictis

With edictis one of the passive participles must be taken after
congruentibus'. accustomed to actions that agree with pronounce
ments'. This makes fair sense, but one expects the simple dictis,
and the construction is unusually involuted, congruens is used
absolutely in 2,7. 25, and factis et dictis may be supported by 2,5
non uerbis modo sed factis etiam securn et cum ceteris congruentem.
4 (i) efficata B, eficata M, aefficata V : effecata F, effaecata edd.:
effecta PL
must have had effecata, since F could hardly have arrived at
it from effecta by conjecture; rather the exemplar of PL corrupted
it into a more familiar word. The choice is basically between
effec-, i.e. effaec-, and effic-. They are both the same word, -ficshowing the vocalization of an old compound, as in caedojeoncido.
It is paralleled by deftcatam in Plautus, Most. 158, and is clearly
the 'difficilior et potior lectio*.
(ii) eiectis codd. (eiectis B): corr. Oudendorp
The qualities named in the following words are certainly not
meant to be cast out, and eiectis can only be defended by positing
a lacuna (see below). In any case, purgation has been dealt with
by the preceding phrase. The subscript points in indicate a
correction, and Oudendorp's minute change gives the required
antithesis; ex animo means of course * whole-heartedly'.
(iii) dehinc scripsi: hinc codd.: hinc <intemperantia, illinc mollitia,
insertis) Novak
Novak's insertion is one of several similar attempts to make
sense at once of eiectis and of hinc. Hinc .. illinc, however, is un
convincing; we would expect the nouns to be joined simply by
et or atque. I propose dehinc, which occurs in 2,2.
5 (i)omnibusque Thomas', obque , ob quae M: absque V
The paradosis is obque \ V has tried to change it into a preposition
suitable for a following ablative. Thomas is probably right to see
in the meaningless word a corruption of oib;q, i.e. omnibusque.
A more plausible correction will hardly be found.
148

(ii) doctrinas F
F did the opposite to V, and accommodated the case of the noun
to the apparent preposition.
6 hanc : primum hoc deinde hec corr. m. rec.
8 (i) repente prius del. Novak '.fort, sapientem
Suddenly' seems out of place in a sentence describing steady
progress along the path of virtue, and the recurrence of the word
immediately in a more fitting place makes it intolerable. Deletion,
however, leaves the final phrase too short for the balance of the
sentence. Sapientem (cf. line 1) is only one possibility.
(ii) hunc Kirchhoff: hoc codd. : hoc est ed. Romana
The neuter pronoun does not make sense as the subject of adtingere, nor can it be taken as an ablative without further alteration.
The editio princeps of 1469, by adding est, makes a facile con
nexion between the infinitives and so bypasses the need for the
subject to be expressed again. But the equivalence implied by
hoc est is contrary to sense; so we need the subject after all, and
hunc (hue) is the obvious correction.
(iii) repetente , in ras.
There is no sense to be seen in this (even with hoc), and it appears
to be a mere error.
(iv) praeteriti F : praeteritis cett.
The genitive is obviously right, and its restoration was well
within the capacity of the copyist of F. It is not clear what caused
the corruption.
9 quod ad modum temporalae (sed modo coeperat exarare): quodam modo intemporale cett. : a ,mr. ed. luntina altera
The view of the sentence's meaning which has led us to adopt
hunc also presupposes intemporalem. After hoc appeared it was
natural for -ale to become -ale. read quoda somehow as quod ad,
and then made modo in into modum so as to have an accusative.
The exemplar apparently lacked word division.
10 uitiis: ut iis
149

A mindless misreading, again involving wrong division.

11 (i) uirtutibus vel bonis add. Thomas : incertisque etiam missis
Purser
The transmitted text is clearly impossible. Purser's conjecture
is palaeographically ingenious, but insertis looks genuine as an
antithesis to exclusis; cf. 1,2 ni Socrates humilitatem cupidinis ex
eius mentihus expulisset et uerae laudis gloriam in eius animum inserere
curasset. It follows that a noun such as uirtutibus is to be added.
(ii) omnibus F
This is another attempt to restore sense to the passage, but a
short-sighted one, as scribes' emendations so often are. It is soon
seen that omnia quae etc. is the subject of pendere.
(iii) fuerunt BMVPL, fuerint F : corr. Stewechius
Even if ad beatam uitam esse were possible Latin, the perfect
tense would be quite out of place. (F is emending again.) Ferunt
is good Latin, and virtually guaranteed by the Platonic passage
that underlies this sentence, Menexenus 247 e
, .
15 (i) ui BVF, post corr. : uis L, ante corr.: tenus
The curious variant in comes from misreading inf a s x u r
or /7 which are found as abbreviations of -tenus. The exemplar
from which L were copied thus had uis, and may have done,
since F could easily have emended to ui. Decision between nomi
native and ablative depends on our judgment of the next variants,
(ii) segregetur : segregere B: segregare V: segregari <(possit)
Novak
The lowest common ancestor of BMV will have had -ere, for it
was natural for this to be made into -are, and it is nearer the reading
of . In other words the agreement of and makes segrege- the
paradosis, and one should not build on the reading of MV as
Novak does, -etur gives excellent sense (with ui) and a good
clausula, while -ere makes no sense. Abbreviations were confused
again, ~ (tur) being read as^~(rtf).
150

17 sed earn quam patientia non Oudendorp : sed (ety earn non putat
quam patientia Rohde: sed patientia Thomas
Oudendorp felt that non enim earn .. sed earn ought to become a
positive statement of what the wise man does regard as contumelia.
But the context implies that there is nothing that provokes him
and that his patience does not tolerate. Rohde's reading makes
supportable insults appear to be a separate category from those
offered by the inprobus, which is awkward. Thomas deletes earn
non putat quam, making it 'the insult which the unrighteous man
offers but which endurance can support'. But it is hard to see why
the words inprobus faciat sed should have been included in the
proposition if it took that form. The transmitted text is better
than any of these attempts. It means, in paraphrase, 'his attitude
is not "I am aggrieved by what is done to me wrongfully" but
rather "I am not aggrieved by what I can tolerate"'.
18 (i) quando Kroll: qua codd.
qua might be defended as meaning according to that law of
nature by which', but quidem does not go well with it, and a simpler
connexion would be more in keeping with the style of the work.
Hence KrolPs excellent emendation. q n = quando, q a = qua.
(ii) scultum
Simplification of the consonant cluster, as in Italian scu/tura.

151

INDEX

Abbreviations 27 f. 89 f. 109.
141.148. 150f.
accents 54 f. 94 f.
ancient critics 18. 19
anthologies 18
apparatus criticus 76. 85 fT.
apographa 12. 33. 68. 86. 118 f.
archetype 32. 38. 41 f. 53
assimilative corruption 23 f.
108f. 111.124. 134. 136.138
asyndeton 22
attribution of speakers 55. 79
author's variants 15f.
banalization 22. 110. 116. 118.
121.128
bibliographies 62ff.73
bowdlerization 18
brackets 80 f. 85
Christian interference 18
closed recension 14. 31 ff.
collating 63 ff.
commentaries, ancient 10. 16;
modern 76
computers 70 ff.
contamination 12f. 35 ff.
criteria of a true reading 48

defending the transmitted text 59

deletions by scribes 81; by
editors 80. 91
diagnostic conjectures 58
dialect forms introduced by
scribes 18f.
dialogue texts 55. 79
difficilior lectio 51. 126. 148
diplomatic transcript 94
dittography 24. 133
division of speakers 55. 79
editor's qualifications 62
elimination of manuscripts 33.
43
emendation 53ff.; by scribes 12.
19f. 22. 32. 50. 107ff. 134f.
139. 143ff. 150
fragments, editions of 76. 95 ff.
frequency of error 119
glosses 22f. 28. 58. 113. 125. 126
grammatical doctrines affect text
19
haplography 24. 139
hiatus in prose 21
homoearchon, homoeoteleuton
25
153

horizontal transmission 142; see

contamination
hyparchetype 33. 75
indexes 98 f. 102
inscriptions 80ff. 94 f.
interpolation 16. 22f. 91. 115.
143 f. 145
itacism 118. 122. 124
lacuna 57. 91 f.
lemma 10. 97
libraries 9. 65
locating corruption 57
manuscripts 9. 30; furnished
with variants 12. 36 f. 42
marginalia 12. 22 f. 28
mental associations cause
corruption 21. 107
metre 61. 82. 137; scribes'
knowledge of it 20. 21. 134.
144
multi-stage corruptions 29. 58
negative apparatus 8714
nomen-sacrum abbreviation
27 f. 124. 141
numeration 76ff. 81. 85. 95 ff.
99 f.
obeli 81
obscure words corrupted 26.144
omissions 24. 42. 80. 91
open recension 14. 37 ff.
orthography 18. 20. 54. 66. 69 f.
86.133.137. 140
154

page-headings 76 f.
palaeographical errors 25 ff. 58 f.
papyri 10. 50. 57. 59. 64. 74f.
80ff.94f. 131 f.
paradosis 53 f.
phonetic errors 20f. 112. 116.
122f. 138
Planudes 19. 2015. 105
positive apparatus 8714
proarchetype 345
proper names corrupted 26;
interpolated 23. 56
prose rhythm 21
punctuation 54f. 57. 69; in the
apparatus 88. 93
quotations 10f. 17f. 78f. 83f.
95 f.
recensions, variant 16f. 70. 75 f.
97
recentiores 50
reminiscence of another passage
causes corruption 21
rubricators 20
saut du meme au meme 24
scholia 10. 17. 82f. 97 f. 107
sigla 73fT. 90
simplification by scribes 22; of
consonant clusters 21. 151. see banalization
spoonerisms 21
statistical methods 46 f.
stemma 14; of variants 52 f. 130

symbols used in editions 80 fF. 93

taxonomy 46 f.
testimonia, see quotations
transpositions 28. 42. 80f. 92.
134
Triclinius 20. 105. 108. 114f.
118
'unnecessary' conjectures 55 f.
unnecessary editions 61

variants, ancient 41 f.; author's

15f.; systematic in certain
manuscripts 19. 125
vitium Byzantinum 21
vox nihili 555
watermarks 30
'weighing' manuscripts 49
word division 26. 54. 149f.
word order subject to variation
21 f. 124. 129. 144

Bibliotheca scriptorum Graecorum et Romanorum

Teubneriana
Epictetl dissertationes (Schenk!)

Inscrlptiones Graecae (SolmsenFraenkel)

Epicuri epistulae (von der Mhll)

Isael orationes (Thalheim)

Firmlci Maternl matheseos libri (Kroll

SkutschZiegler). Cum corr. (Ziegler)
Vol. I. II.
Fragmenta poetarum Latinorum epicorum et
lyricorum praeter Ennium et Lucilium
(Morel)
Fulgentii, Fabii Planciadls, opera. Acc Gordlani Fulgentii de aetatibus mundi et hominis et S. Fulgentii eplscopl super Thebaiden (Helm). Cum. add. (Preaux)
Gal institutiones (SeckelKubier)
Galenl, Claudil, de temperamentis (Helmreich).
Cum add. (Belich)
Gelll noctes Atticae libri XX (Hosius)
Vol. I. II.

Iustlnl epitoma historiarum

Pompel Trogl (Seel). Ed. 2

Philippicarum

Scholia in Iuvenalem vetustiora (Wener)

Llvi, Titl, ab urbe condita libri (WeienbornMller)
Pars III. Libri XXXIXL
Pars IV. Libri XLIXLV. Periochae. Fragmenta. Iulii Obsequentis prodigiorum liber (Rossbach)
Lucanl de bello civili libri X (Hosius). Cum
add. (Balzert). In Vorb.
Adnotationes super Lucanum (Endt)
Scholia in Luclanum (Rabe)
Lydl de magistratibus libri (Wunsch)
de mensibus liber (Wnsch)

Georgli Acropolitae opera (Heisenberg). Ed.

corr. (P. Wirth). In Vorb.
Vol. I. Historia. Breviarium historiae
Vol. II. Scripta minora

Martianus Capella (Dick). Cum add. (Preaux)

Georgli Monadil chronicon (de Boor). Ed. corr.

Mythographl Graecl. Vol. I (Wagner)

(P. Wirth). Vol. I. II. In Vorb.

Grammaticae Romanae fragmenta (Funaioli)
Granu Llclnianl quae supersunt (Flemisch)
Hephaestlonis enchiridion (Consbruch)
Hermogenis opera (Rabe)
Herodlani ab excessu divi Marci libri VIII
(Stavenhagen)
Hesiodl carmina (Rzach)
Hieroclis comm. ad carmen aureum (Khler)
Hyperldls orationes (Jensen)
Iamblldil de communi mathematica scientia
(Festa) Theologumena arithmeticae
(De Falco) In Nicomachi arithmeticam
introductionem (Pistelli). Ed. corr. (U.
Klein). In Vorb.
Iamblldil de vita Pythagorica liber (Deubner). Ed. corr. (U. Klein). In Vorb.
Iamblldil protrepticus (Pistelli)

Melae, Pomponli, de chorographia libri III

(Frick). Cum add. (Schaub)

Pausaniae Graeciae descriptio (Spiro)

Vol. I.III.
Pllni naturalis historiae libri XXXVII (Jan
Mayhoff)
Vol. IV. VI (Indices)
Polyacnl strategematon libri VIII (Woelfflin).
Excerpta. Leonls Imperatorls strategemata
(Melber). Incerti scriptorls Byzantlnl de re
militari liber (Vri). Cum add. (Reinhard)
Polybll historiae (DindorfBttnerWobst)
Vol. IV.
Prolegomenon sylloge (Rabe). Ed. corr.
(Haase). In Vorb.
Pseudacronis scholia in Horatium vetustiora
(Keller)
Vol. I. II.
Qulntilianl declamationes (Ritter)
Qulntl Smyrnael Posthomericorum libri XIV
(Zimmermann). Tryphlodorl et Colluthl
carmina (Weinberger)

Bibliotheca scriptorum Graecorum et Romanorum

Teubneriana
Scriptores metrologicl (Hultsch). Vol. I/II

Valeril Flaccl Argonauticon libri (Kramer)

Senecae (rhet.) oratorum et rhetorum sententiae, divisiones, colores (Kiessling)

Valerli Maxlmi factorum et dictorum memorabilium libri IX (Kempf)

Senecae (phllos.) opera

Vol. II. Naturalium quaestionum libri VIII
(Gercke). Cum add. (Schaub)

Veget epitoma rei militaris (Lang)

Suetoni Tranqullll, C , opera. Vol. I. De vita

Caesarum libri VIII (Ihm)

Vellei Patercull ex historia Romana quae

supersunt (Stegmann v. Pritzwald). Cum
add. (Blume)

Scholia in Theocritum vetera (Wendel)

Xenophontis Athenaion politeia (Kaiinka)

Theodoretl Graecarum
(Raeder)

affectionum

curatio

commentarii (Memorabilia) (Hude)

expeditio Cyri (Anabasis) (Hude)

> Theophylacti Simocattae historiae (de Boor).

Ed. corr. (P. Wirth)

historia Graeca (Hellenica) (Hude)

Sammlung wissenschaftlicher Commentare (SWC)

Catonls praeter librum de re rustica quae exstant (Jordan)
Demetr Phalerel de elocutione libellus (Radermadier)
Dlonysii vel Longini de sublimitate libellus
(JahnVahlen). Cum ind. (Blume)
Eplcurea (Usener)
Hlstorlcorum Romanorum reliquiae (Peter)
Vol. I. Cum add. (Kroymann-Schaub)
Vol. II. Cum add. (Kroymann)
I Itineraria Romana (CuntzSchnetz). Ed. corr.
In Vorb.
Vol. I. II.

Lexlcographl Graeci
Vol. I. Suidae Lexicon (Adler)
Pars 15.
Vol. IX. Pollucis Onomasticon (Bethe)
Fase. 13
Papyri Graecae magicae. Die griechischen
Zauberpapyri. Griech. u. dt. (Preisendanz
et al.). Ed. corr. (Henrichs)
Vol. I. II.
Vol. III. Index. In Vorb.
Sallusti historiarum reliquiae (Maurenbrecher)
Sophokles Elektra. Erkl. v. Kaibel
Stoicorum vetenim fragmenta (v. Arnim)
Vol. IIV.

Griechische und lateinische Schriftsteller (GLS)

Catull. Von Kroll. Add. (HerterKroymann)
Neu

Piaton. Protagoras. Von Nestle. Korr. Ed. (H.

Hofmann). In Vorb.