Hostname: page-component-74656f9884-npj9w Total loading time: 0 Render date: 2026-11-07T01:04:34.051Z Has data issue: false hasContentIssue false
  • English
  • Français

COLUMNAR TRANSLATION: AN ANCIENT INTERPRETIVE TOOL THAT THE ROMANS GAVE THE GREEKS

Published online by Cambridge University Press:  09 June 2015

Eleanor Dickey
Eleanor Dickey*
Affiliation:
University of Reading
*
E.Dickey@reading.ac.uk
Rights & Permissions [Opens in a new window]

Extract

Among the more peculiar literary papyri uncovered in the past century are numerous bilingual texts of Virgil and Cicero, with the Latin original and a Greek translation arranged in distinctive narrow columns. These materials, variously classified as texts with translations or as glossaries, were evidently used by Greek-speaking students when they first started to read Latin literature. They thus provide a unique window into the experience of the first of many groups of non-native Latin speakers to struggle with reading the classics of Latin literature.

Information

Type
Research Article
Information
The Classical Quarterly , Volume 65 , Issue 2 , December 2015 , pp. 807 - 821
Check for updates
Copyright
Copyright © The Classical Association 2015 

Among the more peculiar literary papyri uncovered in the past century are numerous bilingual texts of Virgil and Cicero, with the Latin original and a Greek translation arranged in distinctive narrow columns.Footnote 1 These materials, variously classified as texts with translations or as glossaries, were evidently used by Greek-speaking students when they first started to read Latin literature. They thus provide a unique window into the experience of the first of many groups of non-native Latin speakers to struggle with reading the classics of Latin literature.

Discussion of these papyri has so far focussed on the light they shed on the text of Virgil and Cicero in antiquity, on their use of lectional signs, on codicological issues and on what they reveal about ancient education.Footnote 2 Little attention has been paid to the Greek translations, which in fact are often disregarded as objects of study on the grounds that they are so bad as to be positively painful to read. Not only are they all in prose, but they are very literal, have no literary or stylistic pretensions, and make no attempt to convey the beauty of the original language. Sometimes, moreover, they display serious misunderstandings of the original.

Despite these acknowledged drawbacks, the ancient translations have something important to tell us. They are among the few surviving examples of a system of exegesis that was fundamental to ancient learning and that has generally been overlooked and misunderstood in modern times, in part because it has no parallel in modern teaching or scholarship. If we wish to appreciate how Greek-speaking scholars and students approached Latin literature, an understanding of their unique translation system is essential.

COLUMNAR TRANSLATION: THE BASIC PRINCIPLES

The translation system exemplified in the Virgil and Cicero papyri may be called ‘columnar translation’, because it is based on a system of narrow columns, usually only one to three words wide but capable of containing five or six words per line when necessary.Footnote 3 The Latin is usually in the left-hand column and the Greek in the right-hand column, and each line of the Greek column translates the corresponding line of the Latin column. One can therefore read either across the lines to get a translation of a particular phrase, or down one column to get the complete text in either Greek or Latin. The goal of the translation is not only to make clear the overall meaning of the original, but also to show someone with limited knowledge of the original language how that meaning is achieved, by making it possible to identify which words and phrases of the translation correspond to particular elements of the original. The line breaks are positioned to divide up meaningful units; the translator can use them both to show the reader how the original text is to be construed and to organize groupings that can be successfully translated as a unit.

The columnar translation system works best when the two languages involved are structurally similar to one another. This is the case with Latin and Greek, but less so with either of those languages and English. To illustrate how the system works, therefore, example 1 provides an English columnar translation of a text in a language more closely related to English, namely the opening (lines 354–64) of Goethe's Faust.

  1. 1) 354a Habe nun, I have now,

    b ach! alas!

    c Philosophie, philosophy,

    355a Juristerey law,

    b und Medicin, and medicine,

    356a und leider and unfortunately

    b auch Theologie! also theology

    357a durchaus studirt, thoroughly studied,

    b mit heißem Bemühn. with keen effort.

    358a Da steh’ ich nun, There I stand now,

    b ich armer Thor! poor fool I,

    359a und bin and am

    b so klug as clever

    c als wie zuvor; as before;

    360a Heiße Magister, I am called Master,

    b heiße Doctor gar, am even called Doctor,

    361a und ziehe schon and already I have been leading

    b an die zehen Jahr, for ten years

    362a herauf, herab up, down,

    b und quer and sideways

    c und krumm, and crookedly

    363a meine Schüler my students

    b an der Nase herum – around by the nose –

    364a und sehe, and I see

    b daß wir that we

    c nichts wissen können! cannot know anything!

The English of this translation is not ideal, but it is comprehensible. Because German and English are closely related and have similar grammatical structure, in many lines of this translation the two languages would match no matter where one put the line divisions. But where German and English order differ, the flexibility of the column structure usually makes it possible to produce a translation that matches line for line without doing too much violence to English word order. Thus in lines 358a, 358b, 360b, 361a, 363b and 364c the English words are in a different order from the German ones on the corresponding line, and in lines 354a, 359c, 360a, 360b, 361a, 361b and 364a the English has more or fewer words than the corresponding German.

In antiquity, of course, written texts contained many fewer of the aids that modern readers take for granted. Word division, punctuation, capitalization and diacritical signs such as accents and breathings were only rarely used.Footnote 4 Although the lack of these aids seems to have caused little difficulty for readers familiar with the language in which a text was written, those reading a foreign language would have been handicapped particularly by the lack of word division, which made it difficult even to use a glossary. In verse texts the line breaks normally occurred at the ends of verses, and therefore the reader could at least be sure of finding the beginning of a word at the start of each line, but in prose texts not even that aid was available: columns of prose normally had justified margins, so the line divisions often occurred in the middle of a word, without a hyphen or any other indication that the word had been split between lines. The columnar format would have made life easier for language learners by reducing the number of word divisions they had to locate for themselves: columnar texts only have line divisions at word breaks, and therefore in such texts almost half the word breaks are indicated by line breaks.Footnote 5

Examples 2 and 3 provide two versions of the opening of Goethe's Faust, both of which have been stripped of the aids that an ancient reader would not have had. Example 2 is arranged following the regular layout of poetry in a literary papyrus, with one verse per line.Footnote 6 Example 3 is arranged in the narrow columns associated with columnar translation. Although neither is completely straightforward to read, the second is far easier.

  1. 2) HABENUNACHPHILOSOPHIE

    JURISTEREYUNDMEDICIN

    UNDLEIDERAUCHTHEOLOGIE

    DURCHAUSSTUDIRTMITHEIßEMBEMUHN

    DASTEHICHNUNICHARMERTHOR

    UNDBINSOKLUGALSWIEZUVOR

    HEIßEMAGISTERHEIßEDOCTORGAR

    UNDZIEHESCHONANDIEZEHENJAHR

    HERAUFHERABUNDQUERUNDKRUMM

    MEINESCHULERANDERNASEHERUM

    UNDSEHEDAßWIRNICHTSWISSENKONNEN

  2. 3) HABENUN

    ACH

    PHILOSOPHIE

    JURISTEREY

    UNDMEDICIN

    UNDLEIDER

    AUCHTHEOLOGIE

    DURCHAUSSTUDIRT

    MITHEIßEMBEMUHN

    DASTEHICHNUN

    ICHARMERTHOR

    UNDBIN

    SOKLUG

    ALSWIEZUVOR

    HEIßEMAGISTER

    HEIßEDOCTORGAR

    UNDZIEHESCHON

    ANDIEZEHENJAHR

    HERAUFHERAB

    UNDQUER

    UNDKRUMM

    MEINESCHULER

    ANDERNASEHERUM

    UNDSEHE

    DAßWIR

    NICHTSWISSENKONNEN

The benefits of the columnar system were therefore multiple.

ANCIENT COLUMNAR TRANSLATION UNDER OPTIMUM CONDITIONS

Columnar translation works best in texts that were bilingual from the beginning, because under such circumstances the writer can avoid constructions in either language that would cause difficulties when translated into the other. Of course, the works of Virgil and Cicero were not composed bilingually, but another set of texts for which the columnar format is normally used was indeed so composed: the colloquia of the Hermeneumata Pseudodositheana. The colloquia are a set of little dialogues and narratives designed as easy readers for ancient language learners; the oldest portions seem to have been originally composed for Latin speakers learning Greek and the more recent portions for Greek speakers learning Latin, but all parts of the text appear to have been bilingual from their inception.Footnote 7

Example 4 is an extract from one of the colloquia (Colloquium Montepessulanum 2h), with a third column added in English.

  1. 4) duο ergo sunt δύο οὖν εἰσιν So, there are two

    personae πρόσωπα persons

    quae disputant, τὰ διαλεγόμενα, who converse,

    ego et tu. ἐγὼ καὶ σύ. I and you.

    tu es qui interrogas, σὺ εἶ ὁ ἐπερωτῶν, You are the one who asks;

    ego respondeo. ἐγὼ ἀποκριθήσομαι. I (shall) answer.

The wording is idiomatic in both original languages; it is possible to read either column in isolation and get a perfectly coherent text. The two columns match perfectly line for line (apart from the difference in tense in the last line). But they do not match word for word, because the constructions used are not simply identical: in lines 3 and 5 Greek uses a participle with an article while the Latin uses a relative clause. Latin could not have used the construction employed here in the Greek; Greek could have used the one employed in the Latin, but the participle is more idiomatic. The columnar translation has therefore allowed the writer the freedom to use the most idiomatic construction in each language while still making the two languages correspond closely.

In example 5 (Colloquium Harleianum 1h) the constructions are the same in Latin and Greek, but the word order in the first line is different, and in the last line Greek has an article where Latin does not. Again, therefore, the columnar translation allows both languages to be idiomatic while still making it easy to find the translation of a particular phrase.

  1. 5) si quis autem tibi ἐὰν δέ τίς σοι But if anyone hassles you,

    molestatur, ἐνοχλήσῃ,

    indica μήνυσον tell

    praeceptori. τῷ διδασκάλῳ. the teacher.

In this example the English does not work as well as in the first one, because English requires objects to follow verbs and the placement of the verb on a line by itself after the object makes that impossible without altering the line divisions of the original. This problem, however, arises only because the English has been added after the line divisions were fixed; the original writers did not consider the needs of English translators when dividing up the lines. If we had the same freedom as ancient writers, we could alter the first line division by one word and produce the version in example 6, which would work in all three languages.

  1. 6) si quis autem ἐὰν δέ τίς But if anyone

    tibi molestatur, σοι ἐνοχλήσῃ, hassles you,

    indica μήνυσον tell

    praeceptori. τῷ διδασκάλῳ. the teacher.

In example 7 (Colloquia Monacensia–Einsidlensia 8a) the word order of the Latin and Greek is exactly the same, and the constructions are closely parallel. Nevertheless, the grammar is far from identical: in the fourth line the Greek has a dative and the Latin an ablative, and in the fifth and sixth lines the Greek has a genitive absolute surrounding a dative (as the object of ἀκολουθοῦντος, since ἀκολουθέω takes a dative), while the Latin has an ablative absolute surrounding an accusative (as the object of sequente, since sequor takes an accusative).

  1. 7) paratus ergo ἑτοιμασθεὶς οὖν So having been prepared

    in omnia, εἰς πάντα, for everything,

    processi προῆλθον I went forth

    bono auspicio, καλῇ κληδόνι, with a good omen,

    sequente me ἀκολουθοῦντός μοι followed by my

    paedagogo. παιδαγωγοῦ. paedagogue.

THE COLUMNAR FORMAT COMPARED TO MODERN BILINGUAL FORMATS

Nowadays there are two common formats for bilingual texts. Facing-page translations are generally fairly idiomatic and therefore make the overall meaning of the passage clear, but they often provide little help to the reader who wants to understand exactly what the original text says. Interlinear translations, by contrast, usually tell the reader what the text says but not what it means; it is common for the English of an interlinear translation to make no sense at all when taken as a whole. The contrast is illustrated below in examples 8 and 9, of which the first provides an interlinear translation of the first line of the Iliad and the second a translation that one might find on a facing page.

  1. 8) wrath sing goddess son of Peleus Achilles

    μῆνιν ἄειδε, θεά, ∏ηληϊάδεω Ἀχιλῆος

  2. 9) Sing, goddess, the wrath of Achilles, son of Peleus.

Neither of these systems is entirely satisfactory, for the reader of a bilingual text is very often someone who wants to understand the original language and needs help to do so. Such a person usually needs help both to find out what the text means and to learn what it says, and neither of the usual modern translation systems offers such help. This problem is particularly acute in the field of linguistics, where research frequently involves presenting very specific information about the workings of languages with which readers are largely or even wholly unfamiliar; the writer's entire argument often rests on examples that very few of the readers can understand without help. For this reason linguists usually provide first an interlinear translation in the form of word-by-word glosses containing both lexical and grammatical information, and then a freer translation to give the overall meaning of the sentence. So a linguist might render the first line of the Iliad as in example 10. This solution makes it clear to the reader both what the line means and how and why it means that, but it is very cumbersome: the original five-word line has now acquired twenty-six words of translation and explanation.

  1. 10) μῆνιν ἄειδε θεά

    wrath.acc.sg. sing.imperat.2nd.sg. goddess.voc.

    ∏ηληϊάδεω Ἀχιλῆος

    son of Peleus.gen.sg. Achilles.gen.sg.

    ‘Sing, goddess, the wrath of Achilles, son of Peleus.’

The columnar translation, like the linguists' solution, can be regarded as a compromise between the interlinear and the facing-page systems. Because a line-for-line equivalence offers the translator more flexibility than word-for-word equivalence, it is possible to produce a translation that conveys the meaning of the original. At the same time a columnar translation also gives a language learner a good understanding of what the individual words of the original actually say. As example 11 shows, a columnar translation of the first line of the Iiad offers all the benefits of the facing-page translation together with many of the benefits of the interlinear version, and it does so with only nine words, in contrast to the twenty-six words of the linguists' combined version.

  1. 11) μῆνιν ἄειδε, Sing the wrath,

    θεά, goddess,

    ∏ηληϊάδεω of Peleus' son

    Ἀχιλῆος Achilles

COLUMNAR TRANSLATION IN THE VIRGIL AND CICERO PAPYRI

If we take a fresh look at the bilingual Virgil and Cicero papyri in light of an understanding of the nature and purpose of a columnar translation, those translations suddenly appear far better than they did when implicitly compared to our facing-page translations. Sometimes the Greek is not idiomatic, but this is a small price to pay for a translation that efficiently clarifies both what the original means and what it says. Occasionally the translation is not accurate, but that is a problem with execution rather than principle, and is not surprising if some of the translations were done by learners.

Example 12 comes from a columnar version of Cicero's First Catilinarian,Footnote 8 with the spelling corrected and diacritics added to make the text legible by modern readers. The two versions are essentially the same except in the fourth line, where the Latin gender-neutral parens has been rendered in Greek (which lacks an equivalent gender-neutral term) with μήτηρ; as the word for ‘fatherland’ is feminine in both languages, the use of a feminine word for ‘parent’ is an obvious choice.

  1. 12) nunc te νῦν σε Now of you

    patria ἡ πατρὶς the homeland,

    quae communis est ἥτις κοινή ἐστιν which is the common

    parens μήτηρ mother

    omnium πάντων of all

    nostrum ἡμῶν of us,

    metuit. δέδοικε. has conceived a fear.

Example 13 provides another extract from the same text (section 19 of Cicero's oration). Here the English cannot be made to fit the columnar format completely, but nevertheless the Latin and the Greek work very well; note, in particular, the genitive absolute in Greek corresponding to the Latin quae cum ita sint.

  1. 13) sed quam ἀλλὰ πῶς But how

    longe μακρὰν far away

    uidetur δοκεῖ does it seem that he ought to be

    a carcere ἀπὸ φρουρᾶς from prison

    atque a uinculis καὶ ἀπὸ δεσμῶν and from bonds,

    abesse ἀπεῖναι

    debere ὀφείλειν

    hic qui se οὗτος ὅστις ἑαυτὸν he who himself has judged

    ipse αὐτὸς himself

    iam dignum ἤδη ἄξιον already worthy

    custodia φυλακῆς of confinement?

    iudicauerit? ἔκρινεν;

    quae cum ita sint, τῶν οὕτως ἐχόντων, Since these things are thus,

    Catilina, Κατιλίνα, Catiline,

    debebas ὤφειλες you should have …

THE HISTORY OF COLUMNAR TRANSLATION

Columnar translation probably developed from columnar glossaries, for the format is common for certain types of glossary, and a columnar translation is in effect one that treats a continuous text like a glossary. Columnar glossaries were used in ancient Mesopotamia,Footnote 9 and it is tempting to try to connect the Latin–Greek columnar translations with the Mesopotamian glossaries, but such a connection is unconvincing. The chronological and geographical gaps between the two groups of columnar texts are enormous, for there is clear evidence that Roman Egypt received the columnar translation format from Latin speakers, not from speakers of Greek or Egyptian. It is most unlikely that the Romans would have borrowed anything from the Mesopotamians directly, without going via either of those other cultures. Moreover, the columnar glossary is an idea that two cultures could easily have had independently.

The columnar translation format is by far the most common one for Greek–Latin bilingual papyri (a term that will here be restricted to papyri containing the same material in both languages, excluding those in which the two languages say different things and those in which one language provides only a partial translation of the other, for example via occasional glosses). To illustrate the popularity of the format and the other possibilities available, all the bilingual Greek–Latin papyri whose formats I can ascertain are listed in the table below.Footnote 10 Although our main concern here is with continuous texts rather than with glossaries, all relevant glossaries are included here as well because of their probable role in the development of columnar translations of continuous texts.

Columnar format

Virgil:

  1. 1) P.Ryl. III.478 + P.Mil. I.1 + P.Cairo inv. 85644 A–BFootnote 11 (fourth century a.d., containing parts of Virgil, Aeneid 1 with Greek translation)

  2. 2) BKT IX.39Footnote 12 (fourth century a.d., containing parts of Virgil, Aeneid 1 and 2 with Greek translation)

  3. 3) Ambrosian PalimpsestFootnote 13 (fourth or fifth century a.d., containing parts of Virgil, Aeneid 1 with Greek translation)

  4. 4) P.Fouad 5Footnote 14 (fourth or fifth century a.d., containing parts of Virgil, Aeneid 3 with Greek translation)

  5. 5) P.Oxy. L.3553Footnote 15 (fifth century a.d., containing parts of Virgil, Aeneid 1 with Greek translation)

  6. 6) P.Vindob. inv. L 24Footnote 16 (fifth century a.d., containing parts of Virgil, Aeneid 5 with Greek translation)

  7. 7) A papyrus edited originally by HusselmanFootnote 17 (fifth century a.d., containing parts of Virgil, Georgics 1 with Greek translation)

  8. 8) P.Ness. II.1Footnote 18 (sixth century a.d., containing portions of Virgil, Aeneid 1 and 2 with Greek translation)

  9. 9) P.Vindob. inv. L 62Footnote 19 (sixth century a.d., containing parts of Virgil, Aeneid 2 with Greek translation)

Cicero:

  1. 10) P.Rain.Cent. 163Footnote 20 (fourth or fifth century a.d., containing parts of Cicero, In Catilinam 1 with Greek translation)

  2. 11) PSI Congr.XXI 2Footnote 21 (fifth century a.d., containing parts of Cicero, In Catilinam 1 with Greek translation)

  3. 12) P.Ryl. I.61Footnote 22 (fifth century a.d., containing parts of Cicero, In Catilinam 2 with Greek translation)

  4. 13) P.Vindob. inv. L 127Footnote 23 (fifth century a.d., containing parts of Cicero, In Catilinam 3 with Greek translation)

Colloquia:

  1. 14) P.Berol. inv. 21860Footnote 24 (fourth century a.d., containing phrases from an otherwise unknown bilingual colloquium mixed with glossary material)

  2. 15) P.Prag. II.118Footnote 25 (fourth or fifth century a.d., containing a bilingual colloquium closely related to the Colloquium Harleianum)

Other continuous texts:

  1. 16) BKT IX.149Footnote 26 (fourth century a.d., containing Isocrates with Latin translation)

  2. 17) PSI VII.848Footnote 27 (fourth century a.d., containing Aesop fable 264 with Latin translation; format not quite certain owing to small size of surviving fragment)

  3. 18) P.Bon. 5Footnote 28 (third or fourth century a.d., containing model epistles in Latin and Greek)

  4. 19) CLA Footnote 29 II.251 (sixth or seventh century a.d., containing part of the Bible with Latin translation)

Glossaries:Footnote 30

  1. 20) P.Oxy. LXXVIII.5162 (first or second century a.d.)

  2. 21) P.Oxy. LXXVIII.5163 (first or second century a.d.)

  3. 22) P.Oxy. XLIX.3452Footnote 31 (second century a.d.)

  4. 23) P.Lund I.5Footnote 32 (second century a.d.)

  5. 24) Kramer (n. 28), no. 12Footnote 33 (second or third century a.d.)

  6. 25) P.Oxy. XXXIII.2660aFootnote 34 (third century a.d.)

  7. 26) P.Laur. IV.147Footnote 35 (third century a.d.)

  8. 27) Kramer (n. 24), no. 4Footnote 36 (third or fourth century a.d.)

  9. 28) Kramer (n. 24), no. 6Footnote 37 (third or fourth century a.d.)

  10. 29) Kramer (n. 24), no. 3Footnote 38 (third or fourth century a.d.)

  11. 30) P.Oxy. LXXVIII.5161 (third or fourth century a.d.)

  12. 31) Kramer (n. 28), no. 10Footnote 39 (fourth century a.d.)

  13. 32) P.Fay. 135v descr.Footnote 40 (fourth century a.d.)

  14. 33) P.Lond. II.481Footnote 41 (fourth century a.d.)

  15. 34) PSI VII.756Footnote 42 (fourth or fifth century a.d.)

  16. 35) P.Oxy. VIII.1099Footnote 43 (fifth century a.d.)

  17. 36) Fragmenta Helmstadiensia + Folium WallraffianumFootnote 44 (sixth century a.d.)

Facing-page format

  1. 37) PSI XIII.1306 (LDAB 3024, fourth or fifth century a.d., containing parts of the Bible with Latin translation): format is not completely certain because of the small size of the fragment, but probably facing pages with Greek on the left.

  2. 38) Codex Bezae (LDAB 2929, fifth century a.d., containing parts of the Bible with Latin translation): Greek on the left

  3. 39) CLA (n. 29), V.521 (LDAB 3003, sixth century a.d., containing parts of the Bible with Latin translation): Greek on the left

  4. 40) CLA (n. 29), IV.472 (LDAB 3344, sixth or seventh century a.d., containing parts of the Bible with Latin translation; the Greek is in the Latin alphabet): Greek on the left

  5. 41) CLA (n. 29), V.520 (LDAB 3403, seventh century a.d., containing parts of the Bible with Latin translation): Latin on the left

The translation follows the original in the same column

  1. 42) P.Yale II.104 + P.Mich. VII.457Footnote 45 (third century a.d., containing Aesop with Latin translation)

  2. 43) P.Amh. II.26Footnote 46 (third or fourth century a.d., containing Babrius with Latin translation)

Other formats

  1. 44) BKT IX.150Footnote 47 (first century b.c., containing a glossary): a single column, in which each Latin gloss is underneath the corresponding Greek lemma and slightly indented.

  2. 45) P.Sorb. inv. 2069 versoFootnote 48 (third century a.d., containing glossary with grammatical explanations in continuous text): an originally columnar text has been copied in long lines, so that short Latin and Greek phrases alternate; the languages are divided by spaces, and new lemmata do not necessarily begin new lines.Footnote 49

  3. 46) Chester Beatty codex AC 1499Footnote 50 (fourth century a.d., containing among other things a glossary to the Pauline epistles): in the glossary section each Greek word is followed by its Latin translation(s), with double points separating lemma from gloss and multiple glosses from each other, while a unique symbol like a modern double quotation mark (“) separates the different entries. Line breaks are irrelevant to the arrangement of the text and often occur in the midst of words. This format may, but does not have to, result from re-arranging a text that originally used the columnar layout.

  4. 47) P.Berol. inv. 10582Footnote 51 (fifth or sixth century a.d., containing a trilingual colloquium in Latin, Greek and Coptic): an originally columnar text has been put into the usual format for Coptic glossaries (see below) by replacing the intercolumnar spaces with double points, so that each line has three short units (one in each language) separated by punctuation.Footnote 52

Thus the evidence consists of thirty-six columnar papyri (nineteen containing continuous text and seventeen glossaries) and eleven others (nine of which contain continuous text). The distribution of material into these two categories is not random: when a continuous literary text originally composed in one language has been provided with a translation in the other language, the format is always columnar if the original language was Latin, and usually non-columnar if the original language was Greek. Within this latter group there appear to be subdivisions connecting genre and format, for facing-page translations are used only for Biblical texts and translations that follow the original only for fables.

The apparent connection between an originally Latin text and columnar format is reinforced by the fact that papyri not containing Latin almost never use this format. Of course, monolingual Greek papyri by definition do not contain translations of continuous text, but we have numerous Greek–Greek glossaries (mostly Homer lexica, but occasionally lexica of other types), and these normally use a format in which the gloss follows immediately after the lemma, separated by a space (or sometimes by punctuation, or occasionally not separated at all) rather than by the start of a new column. If the gloss is longer than average, it usually continues on a second line, which begins under the lemma but slightly indented. Of the thirty-nine Greek–Greek glossaries whose formats I have been able to verify, thirty-four use this format,Footnote 53 three a different non-columnar formatFootnote 54 and only two the columnar format.Footnote 55

Bilingual Greek–Demotic and Greek–Coptic texts seem never to use the columnar format at all, at least not during antiquity.Footnote 56 I can find only one bilingual Greek–Demotic text, a glossary, and this uses the same format as the majority of the Greek–Greek glossaries.Footnote 57 Greek–Coptic glossaries also use this format, the only difference being that, whereas Greek–Greek glossaries usually have a space after the lemma, or failing that a high point, Greek–Coptic glossaries tend to divide the lemma from the gloss with a double point (like a modern colon).Footnote 58 Greek–Coptic continuous bilingual texts use a variety of formats, of which the most common during antiquityFootnote 59 is for the translation to follow the text in the same column;Footnote 60 other formats include having the text on one side of a page and the translation on the other,Footnote 61 the facing-page format,Footnote 62 and parallel columns in which the two languages do not match line for line.Footnote 63

The obvious inference from the connection between Latin language and columnar format is that the columnar translation format originated in the Latin-speaking areas of the empire. Latin speakers had been learning Greek for centuries before Greek speakers began to learn Latin on any comprehensive scale;Footnote 64 therefore, it is inherently likely that some of the Latin–Greek bilingual materials (especially glossaries and colloquia) originated in the West for use by Latin speakers and were later adapted for use by Greek speakers. Some texts show positive evidence of a Western origin and later Eastern adaptation.Footnote 65 If the materials themselves migrated across the empire, it is not surprising that their format came with them.

The colloquia are among the materials that probably originated in the West, and it is notable that they are universally found in columnar format, not only in papyri but also in medieval manuscripts; only in the Renaissance do colloquium manuscripts with other formats start to appear. But the bilingual texts of Virgil and Cicero cannot have originated in the West: those are clearly designed for Greek speakers learning Latin. The first teachers who produced such texts were probably expatriate Latin speakers teaching Greek in the East; they would have used the columnar format they knew and appreciated from their own studies to help their students with Latin texts.

Our understanding of the mechanics of teaching and scholarship in the ancient West is limited, especially in comparison with the vast resources the papyri provide for understanding the education system of the Greek East.Footnote 66 Apart from a few rather sparse descriptions in literary texts, all we can do to understand what sort of materials teachers, students and scholars used in the Western empire is to extrapolate from the materials we have from Greek-speaking Egypt. Given the Romans' respect for Greek literature, culture and scholarship, the traditional assumption that Roman education was modelled largely on Greek education has not been an unreasonable one. But in the case of columnar translation the influence seems to have gone the other way: a technique developed in the West was borrowed by teachers in the East.

If this technique had not happened to involve Greek as well as Latin, it would not have been borrowed by people living in a climate that preserves writing materials, and we would not now know about it at all. Under these circumstances it is perhaps worth considering whether there are other respects in which Western education may have been less similar to that in the East than we normally suppose.

References

1 I am grateful to Roger Bagnall, Daniela Colomo, Martin West, Philomen Probert, Rolando Ferri and CQ's anonymous but extremely knowledgeable reader for their help with this project. I am also grateful to Serena Ammirati and Marco Fressura for sharing their unpublished work on the layout of bilingual texts with me (S. Ammirati and M. Fressura, ‘Towards a typology of ancient bilingual glossaries: palaeography, bibliology, and codicology’, forthcoming in T. Derda, J. Urbanik, A. Łajtar and G. Ochała, Proceedings of the XXVII International Congress of Papyrology [Warsaw]); they independently make some of the points that are made below, as well as discussing other aspects of layout (e.g. indentation, use of paragraphoi).

2 See e.g. Gaebel, R.E., ‘The Greek word-lists to Vergil and Cicero’, Bulletin of the John Rylands Library 52 (1970), 284–325 CrossRefGoogle Scholar; Scappaticcio, M.C., Papyri Vergilianae: l'apporto della papirologia alla storia della tradizione virgiliana (I–VI d.C.) (Liège, 2013)Google Scholar; Rochette, B., ‘Les traductions grecques de l'Énéide sur papyrus: une contribution à l’étude du bilinguisme gréco-romain au Bas-Empire’, Les Études Classiques 58 (1990), 333–46Google Scholar; id., Le latin dans le monde grec (Brussels, 1997)Google Scholar, esp. 302–15; Fressura, M., ‘Tipologie del glossario virgiliano’, in Marganne, M.-H. and Rochette, B. (edd.), Bilinguisme et digraphisme dans le monde gréco-romain: l'apport des papyrus latins (Liège, 2013), 71–116 Google Scholar; Axer, J., ‘Reedition of the Viennese fragments of Cicero, In Catilinam I’, in Festschrift zum 100-jährigen Bestehen der Papyrussammlung der Österreichischen Nationalbibliothek, Papyrus Erzherzog Rainer (Vienna, 1983), 468–82Google Scholar; Maehler, H., ‘Zweisprachiger Aeneis-codex’, in Bingen, J. and Nachtergael, G. (edd.), Actes du XVe congrès international de papyrologie II: Papyrus inédits (Brussels, 1979), 18–41 Google Scholar; Reichmann, V., Römische Literatur in griechischer Übersetzung (Leipzig, 1943), 28–57 Google Scholar.

3 How much narrower this is than the columns of monolingual papyri depends on the genre. Columnar papyri of Virgil take on average four to five lines to cover one hexameter (see the editions in Scappaticcio [n. 2]), and therefore the average line length is less than one-quarter of the line length in a monolingual text of Virgil. But with Cicero the difference is less great, because while the columns in bilingual papyri are the same width for any genre, columns in monolingual papyri are narrower for prose than for hexameters: according to Johnson, W.A., Bookrolls and Scribes in Oxyrhynchus (Toronto, 2004)Google Scholar, 101 and 116, the usual width of a column of hexameter verse is 10.4–13.6 cm, whereas that for a column of prose is 4.3–7.5 cm. Nevertheless, the columns in bilingual texts of Cicero are still narrower than those in monolingual texts, for the average width of a bilingual column of Cicero ranges from 3 to 4.7 cm (according to D. Internullo, ‘Cicerone latinogreco: corpus dei papiri bilingui delle Catilinarie di Cicerone’, Papyrologica Lupiensia 20–21 [2011–12], 25–150, at 38, 80, 95 and 108, the average column width is 3 cm in P.Rain.Cent. 163, 3.5 cm in P.Ryl. 1.61, 4 cm in PSI Congr.XXI 2 and 4.7 cm in P.Vindob. L 127).

4 See for example the plates in Turner, E. and Parsons, P., Greek Manuscripts of the Ancient World (London, 1987 2)Google Scholar, with discussion on pp. 8–12 of such aids as do occur; for diacritics see also e.g. Nodar, A., ‘Ancient Homeric scholarship and the medieval tradition: evidence from the diacritics in the papyri’, in Palme, B. (ed.), Akten des 23. internationalen Papyrologenkongresses (Vienna, 2007), 469–81Google Scholar; for word dividers also e.g. E. Dickey, ‘Word division in bilingual texts', in G.N. Macedo and M.C. Scappaticcio (edd.), Signes dans les textes et textes sur les signes (Liège, forthcoming).

5 Cf. the discussion by Hock, R.F. and O'Neil, E.N., The Chreia in Ancient Rhetoric: Classroom Exercises (Leiden, 2002), 7–8 Google Scholar, of P.Bour. 1.141–70 (= M–P 3 2643, LDAB 2744), a fourth-century monolingual Greek papyrus that uses columnar format to present reading material for children first progressing from isolated words to connected sentences, and then moves to longer lines as the student advances. The layout of this papyrus seems to be unique; nevertheless, its existence demonstrates that someone found the columnar layout useful for children first learning to read.

6 The line length would have been greater in antiquity, for Goethe's verses are shorter than those of the hexameter poetry typically read by ancient language learners (Quintilian, Inst. 1.8.5 tells us that Latin speakers started their Greek reading with Homer, and the papyri tell us that Greek speakers learning Latin started with Virgil and moved on to Terence, Juvenal and Seneca: see Dickey, E., The Colloquia of the Hermeneumata Pseudodositheana [Cambridge, 2012–15], 1.7–10CrossRefGoogle Scholar).

7 For the origins and development of the colloquia, which are complex, see the introduction to Dickey (n. 6). Quotations from the colloquia and references to them are hereafter given according to that edition; if the letter at the end is subtracted, the same references can be used to find the passage concerned in the appendix of Goetz's edition ( Goetz, G., Hermeneumata pseudodositheana; vol. 3 of Corpus Glossariorum Latinorum [Leipzig, 1892]Google Scholar). In most medieval copies of the colloquia the Greek occupies the left-hand column and the Latin the right, but I have reversed that order here because the papyrus evidence suggests that the ancient copies normally had the Latin on the left and the Greek on the right.

8 P.Rain.Cent. 163, edited by Internullo (n. 3), 37–79 (= M–P 3 2922, LDAB 554), fol. Iv, lines 33–9; the lines quoted here come from section 17 of the speech.

9 For examples see Nougayrol, J., Laroche, E., Virolleaud, C. and Schaeffer, C.F.A., Ugaritica V (Paris, 1968), 230–49Google Scholar; for discussion see Civil, M., ‘Ancient Mesopotamian lexicography’, in Sasson, J.M. (ed. in chief), Civilizations of the Ancient Near East (New York, 2000), 4.2305–14Google Scholar.

10 For papyrological abbreviations see the Checklist at http://library.duke.edu/rubenstein/scriptorium/papyrus/texts/clist.html; further information on each text listed here can be found in the databases referred to (M–P 3 = Mertens–Pack database, http://promethee.philo.ulg.ac.be/cedopal/indexsimple.asp; LDAB = Leuven Database of Ancient Books, http://www.trismegistos.org/ldab/).

11 Edited by Scappaticcio (n. 2), no. 5; M–P 3 2940; LDAB 4146.

12 Edited by Scappaticcio (n. 2), no. 4; M–P 3 2939.1; LDAB 4149.

13 Edited by Scappaticcio, M.C., ‘Appunti per una riedizione dei frammenti del palinsesto Virgiliano dell'Ambrosiana’, Archiv für Papyrusforschung 55 (2009), 96–120 Google Scholar, and Scappaticcio (n. 2), no. 8; M–P 3 2943; LDAB 4156.

14 Edited by Scappaticcio (n. 2), no. 15; M–P 3 2948; LDAB 4154.

15 Edited by Scappaticcio (n. 2), no. 9; Fressura, M., ‘Revisione di POxy VIII 1099 e POxy L 3553’, Studi di Egittologia e di Papirologia 6 (2009), 43–71 Google Scholar; M–P 3 2943.1; LDAB 4160.

16 Edited by Scappaticcio (n. 2), no. 20; M–P 3 2951; LDAB 4161.

17 Husselman, E.M., ‘A palimpsest fragment from Egypt’, in Studi in onore di Aristide Calderini e Roberto Paribeni (Milan, 1957), 2.453–9Google Scholar; Scappaticcio (n. 2), no. 33; M–P 3 2936; LDAB 4159.

18 Edited by Scappaticcio (n. 2), no. 6; M–P 3 2939; LDAB 4166. This papyrus also contains a glossary to portions of Book 4 (i.e. selected words only, but in the inflected forms and in the order that they appear in Virgil's text); evidently the students for whom the papyrus was designed were supposed to be able to progress from using a full translation to using such a glossary by the time they got to Book 4. Fressura (n. 2), 86 has argued that this shift at the start of Book 4 was standard in the teaching of Virgil to Greek speakers.

19 Edited by Scappaticcio (n. 2), no. 11; M. Fressura, ‘PVindob L 62 identificato’, ZPE 168 (2009), 83–96; M–P 3 2944.1; LDAB 6194.

20 Edited by Internullo (n. 3), no. I; M–P 3 2922; LDAB 554.

21 Edited by Internullo (n. 3), no. IV; M–P 3 2921.01; LDAB 556.

22 Edited by Internullo (n. 3), no. II; M–P3 2923; LDAB 4135.

23 Edited by Internullo (n. 3), no. III; M–P3 2923.1; LDAB 559.

24 Edited as continuous text by Kramer, J., Glossaria bilinguia altera (Munich, 2001)Google Scholar, no. 9; new edition in which the material is argued to be less coherent in Dickey (n. 6) vol. 2 section 4.2; M–P 3 3004.02; LDAB 8897.

25 Edited by Dickey, E. and Ferri, R., ‘A new edition of the Colloquium Harleianum fragment in P.Prag. 2.118’, ZPE 180 (2012), 127–32Google Scholar; M–P 3 3004.22; LDAB 6007.

26 CPF 1.2.2 21 116 T & 119 T; M–P 3 1251.02; LDAB 2528.

27 Edited by Kramer (n. 24), no. 10; M–P 3 52; LDAB 138.

28 Edited by Kramer, J., Glossaria bilinguia in papyris et membranis reperta (Bonn, 1983)Google Scholar, no. 16; M–P 3 2117; LDAB 5498.

29 Lowe, E.A., Codices Latini Antiquiores (Oxford, 1934–71)Google Scholar = LDAB 2881.

30 Four other papyri probably belong in this section but are too fragmentary for their format to be ascertained with certainty: P.Oxy. XXXIII.2660 (= Kramer [n. 28], no. 6; M–P 3 2134.1; LDAB 4497; first or second century a.d.), P.Oxy. XLVI.3315 (= Kramer [n. 28], no. 8; M–P 3 3004.2; LDAB 4498; first or second century a.d.), P.Sorb. I.8 (= Kramer [n. 28], no. 3; M–P 3 3008; LDAB 5439; third century a.d.), and P.Vindob. inv. L 150 (= Kramer [n. 24], no. 5; M–P 3 2134.6; LDAB 6053; fifth century a.d.).

31 Edited by Kramer (n. 24), no. 7; M–P 3 2134.7; LDAB 4812.

32 Edited by Kramer (n. 28), no. 9; M–P 3 3004; LDAB 4741.

33 M–P 3 2685.1; LDAB 5062.

34 Edited by Kramer (n. 28), no. 7; M–P 3 2134.2; LDAB 5382.

35 Edited by Kramer (n. 28), no. 5; M–P 3 2134.3; LDAB 4675.

36 M–P 3 3004.21; LDAB 5755.

37 M–P 3 2134.61; LDAB 9218.

38 M–P 3 2134.71; LDAB 9217.

39 M–P 3 3007; LDAB 5631.

40 Edited by Kramer (n. 28), no. 11; M–P 3 2013.1; LDAB 7680.

41 Edited by Kramer (n. 28), no. 13; M–P 3 3005; LDAB 5678.

42 Edited by Scappaticcio (n. 2), no. 13; M–P 3 2946; LDAB 4155.

43 Edited by Scappaticcio (n. 2), no. 19; Fressura (n. 15); M–P 3 2950; LDAB 4162.

44 Edited by Kramer (n. 28), no. 4; M–P 3 2134.4; LDAB 6279.

45 M–P 3 2917; LDAB 134.

46 Edited by Kramer, J., Vulgärlateinische Alltagsdokumente auf Papyri, Ostraka, Täfelchen und Inschriften (Berlin, 2007)CrossRefGoogle Scholar, no. 10; M–P 3 172; LDAB 434.

47 Edited by Kramer (n. 28), no. 1; M–P 3 2134.5; LDAB 6764.

48 Edited by Dickey, E. and Ferri, R., ‘A new edition of the Latin–Greek glossary on P.Sorb. inv. 2069 (verso)’, ZPE 175 (2010), 177–87Google Scholar; M–P 3 3006; LDAB 5438.

49 See Dickey, E., ‘The creation of Latin teaching materials in antiquity: a re-interpretation of P.Sorb. inv. 2069’, ZPE 175 (2010), 188–208 Google Scholar.

50 Edited by Wouters, A., The Chester Beatty Codex AC 1499: A Graeco-Latin Lexicon on the Pauline Epistles and a Greek Grammar (Leuven 1988), 115–47Google Scholar for the glossary; M–P 3 2161.1; LDAB 3030.

51 Edited by Dickey, E., ‘How Coptic speakers learned Latin? A reconsideration of P.Berol. inv. 10582’, ZPE 193 (2015), 65–77 Google Scholar.

52 For further information on this papyrus and its layout see Dickey (n. 51).

53 As this is a large group I give only the LDAB numbers, in chronological order from third/second century b.c. to sixth century a.d.: LDAB 2344, 7028, 1330, 1460, 9945, 1566, 1634, 1640, 1659, 1712, 1729, 1854, 4558, 4560, 4806, 1674, 1817, 1830, 1841, 5091, 1948, 1969, 1987, 2016, 2022, 2023, 109068, 2060, 2063, 1689, 2118, 10228, 2208, 6322. A number of these are laid out in columns in modern editions, but I have verified the original format from photographs.

54 Henrichs, A., ‘Scholia minora zu Homer III’, ZPE 8 (1971), 1–12 Google Scholar, no. 9 (= M–P 3 1209.5, LDAB 1456, first or second century a.d.); Henrichs, A., ‘Scholia minora zu Homer II’, ZPE 7 (1971), 229–60Google Scholar, no. 4 (M–P 3 1166, LDAB 1516, second century a.d.); P.Oslo II.12 (M–P 3 1160, LDAB 1669, second century a.d.).

55 Gallazzi, C., ‘Glossario a Homerus, Odyssea I 46–53’, ZPE 45 (1982), 41–6Google Scholar (= M–P 3 1207.1, LDAB 1390, first century a.d.) and PapCongr. XX p. 285 no. 3 (M–P 3 1163.01, LDAB 2252, seventh century a.d.). The format of the second of these is not quite certain; it looks columnar to me from the photograph, but only three lines are well preserved, and traces of a fourth have led the editor to believe that it did not line up with the other three, making the glossary non-columnar.

56 The basis of this statement is a search (on 31 January 2013) of the Leuven database for papyri containing both Greek and Coptic or Demotic, followed by inspection of editions of all the resulting papyri dating to the sixth century a.d. or earlier, at least in so far as those editions could be located in the Sackler and Bodleian Libraries in Oxford. Most papyri consulted proved not to be bilingual as defined for the purposes of this article; for those that were indeed bilingual I then consulted photographs to verify the original layout, unless this was specified in the literature, since the layouts of editions do not always match those of the originals (editors have a tendency to separate Coptic-style glossaries into columns to make them easier to read). See also the detailed discussion of layout of Greek–Coptic bilingual papyri of the Old Testament by Nagel, who does not mention the columnar format ( Nagel, P., ‘Griechisch–koptische Bilinguen des Alten Testaments’, in id. [ed.], Graeco–Coptica: Griechen und Kopten im byzantinischen Ägypten [Halle, 1984], 231–57Google Scholar).

57 Quecke, H., ‘Eine griechisch–ägyptische Wörterliste vermutlich des 3. Jh. v. Chr. (P. Heid. Inv.-Nr. G 414)’, ZPE 116 (1997), 67–80 Google Scholar (= M–P 3 2131.02, LDAB 6962, third century b.c.).

58 e.g. P.Rain.UnterrichtKopt. 257a (LDAB 3141, third or fourth century a.d.); P.Rain.Cent. 12 (= M–P 3 2133.2, LDAB 6614, seventh century a.d.); P.Rain.UnterrichtKopt. 280 (= M–P 3 2698, LDAB 6668, seventh or eighth century a.d.); P.Rain.UnterrichtKopt 264 (= LDAB 10974, undated); SB Kopt III.1656 (= M–P 3 2132, LDAB 5647, fourth century a.d.; the format of this glossary is not quite like that of the others, but it is certainly not columnar).

59 Greek–Coptic bilingual texts continued to be produced throughout the medieval period, and in fact the majority of those listed on the LDAB are medieval. Because medieval developments are not relevant to the question of origin investigated here, I have only looked at continuous Coptic texts datable to the sixth century a.d. or earlier.

60 e.g. Schmidt, C. and Schubart, W., Acta Pauli nach dem Papyrus der Hamburger Staats- und Universitäts-Bibliothek (Hamburg, 1936)Google Scholar (= LDAB 3138, third or fourth century a.d., parts of the Bible with Coptic translation); Amundsen, L., ‘Christian papyri from the Oslo collection’, Symbolae Osloenses 24 (1945), 121–47CrossRefGoogle Scholar, at 121–40 (= LDAB 2993, fourth century a.d., parts of the Bible with Coptic translation); Rösch, F., Bruchstücke des ersten Clemensbriefes, nach dem achmimischen Papyrus der Strassburger Universitäts- und Landesbibliothek (Strasbourg, 1910), 119–22Google Scholar = LDAB 2806, fifth century a.d., parts of the Bible with Coptic translation); P.Köln IV.169 (= LDAB 3238, fifth century a.d., parts of the Bible with Coptic translation); Codex Scheide (ed. Schenke, H.-M., Das Matthäus-Evangelium im mittelägyptischen Dialekt des Koptischen [Berlin, 1981]Google Scholar = LDAB 107734, fifth century a.d., doxology with Coptic translation); Römer, C., ‘Das zweisprachige Archiv aus der Sammlung Flinders Petrie’, ZPE 164 (2008), 53–62 Google Scholar, at 61–2, no. 26 (= LDAB 10092, fifth or sixth century a.d., early Christian text with Coptic translation); P.Rain.UnterrichtKopt 269 II (= LDAB 2719, fifth or sixth century a.d., Menandri Sententiae with Coptic translation); P.Rain.UnterrichtKopt 269 I (= M–P 3 1583, LDAB 2452, fifth to seventh century a.d., Menandri Sententiae with Coptic translation); P.Rain.UnterrichtKopt 268 (= M–P 3 1583.2, LDAB 2723, sixth or seventh century a.d., one of the Menandri Sententiae with Coptic translation); Biblia Coptica I.III Sa 72 (= LDAB 3195, sixth or seventh century a.d., parts of the Bible with Coptic translation); Biblia Coptica IV.III Sa 700 (= LDAB 2897, parts of the Bible with Coptic translation).

61 e.g. Brashear, W.M. and Satzinger, H., ‘Ein akrostichischer griechischer Hymnus mit koptischer Übersetzung (Wagner-Museum K 1003)’, Journal of Coptic Studies 1 (1990), 37–58 CrossRefGoogle Scholar (= LDAB 5584, third or fourth century a.d., Greek hymn with Coptic translation); Brashear, W.M. and Quecke, H., ‘Ein Holzbrett mit zweisprachigen Hymnen auf Christus und Maria’, Enchoria 17 (1990), 1–19 Google Scholar (= LDAB 5943, fifth century a.d., Greek hymn with Coptic translation).

62 e.g. Pintaudi, R., Antinoupolis (Florence, 2008), 146–7Google Scholar, no. 6 (= LDAB 113257, fifth century a.d., Biblical); Till, W.C. and Sanz, P., Eine griechisch–koptische Odenhandschrift (Rome, 1939)Google Scholar (= LDAB 3483, sixth century a.d., Biblical); Treu, K., ‘Griechisch–koptische Bilinguen des Neuen Testaments’, Koptologische Studien in der DDR (Halle, 1965), 95–123 Google Scholar, at 111–13 (= LDAB 2898, sixth century a.d., Biblical).

63 e.g. Elanskaya, A.I., The Literary Coptic Manuscripts in the A.S. Pushkin State Fine Arts Museum in Moscow (Leiden, 1994), 458–60CrossRefGoogle Scholar (= LDAB 2866, fourth or fifth century a.d., Biblical) and Wessely, K., ‘Ein faijumisch–griechisches Evangelienfragment’, Vienna Oriental Journal 26 (1912), 270–4Google Scholar (= LDAB 2965, sixth century a.d., Biblical). Probably also to be put in this category are two papyri of which only one column survives and whose layout cannot therefore be completely verified: MPER NS 9 pp. 49–51 no. 3 (= LDAB 2964, sixth century a.d., Biblical) and Treu (n. 62), 100–4 (= LDAB 2815, sixth century a.d., Biblical).

64 Already in the Republic it was normal for educated Latin speakers to have studied Greek. Exactly when significant numbers of Greek speakers began learning Latin probably varied from province to province, as some Greek-speaking areas came under Roman domination centuries before others. But in Egypt significant Latin learning seems to have begun in the second century a.d., to judge from the dates of preserved Latin–Greek glossaries (see examples 19 and following in the list above) and from the dates at which Latin loanwords start to be used in Greek papyri (see Dickey, E., ‘Latin influence on the Greek of documentary papyri: an analysis of its chronological distribution’, ZPE 145 [2003], 249–57Google Scholar, at 252). For further information on the learning of Latin by Greek speakers see Rochette (n. 2 [1997]).

65 See Dickey (n. 49) and Dickey (n. 6), 1.39–52.

66 See Bonner, S.F., Education in Ancient Rome: From the Elder Cato to the Younger Pliny (London, 1977), 165Google Scholar, and note the concentration on Eastern evidence in Cribiore, R., Writing, Teachers, and Students in Graeco–Roman Egypt (Atlanta, 1996)CrossRefGoogle Scholar; Cribiore, R., Gymnastics of the Mind: Greek Education in Hellenistic and Roman Egypt (Princeton, 2001)CrossRefGoogle Scholar; and Morgan, T., Literate Education in the Hellenistic and Roman Worlds (Cambridge, 1998)Google Scholar.