The oldest extant papyrus fragment of the Gospel according to Matthew consists of five
small scraps, three of which are kept at Magdalen College, Oxford, the other two at the
Fundación San Lucas Evangelista, Barcelona. It was dated, by Colin Roberts, to the later second
century.2 Roberts himself was the first scholar to recognize the relationship between the three
Magdalen scraps and the two remnants in Barcelona (P.Barc. inv. 1, ∏67) as parts of one and the
same original codex.3 Further attempts to link this codex to fragments of Luke's gospel preserved
at the Bibliothèque Nationale, Paris (Suppl. Gr. 1120 = Gregory-Aland ∏ 4 ), had to be
abandoned:4 although the fragmentary codex at the Bibliothèque Nationale had, at one stage,
contained Matthew – as seems to be obvious from a scrap with the title EÈagg°lion katå
Mayya›on –, the Paris codex is written on much darker, brownish papyrus and is considerably
later (by up to one-hundred years). As yet, there is no candidate among extant papyri to
supplement ∏64 and ∏67. However, after more than forty years since Roberts first published the
Magdalen fragment, some additions and corrections appear to be called for.
Whereas the earliest publications of and about the fragments do not give them a college
library number, van Haelst's Catalogue,5 faithfully copied by all later publications including the
latest, 27th edition of Nestle-Aland, Novum Testamentum Graece, calls them "Gr.18". There are,
however, no such fragments at Magdalen College Oxford. The College handlist had indeed
numbered the papyrus "Magdalen Greek 18", but this was an obvious mistake due to a tiny scrap
1H.C. Youtie, The Textual Criticism of Documentary Papyri. Prolegomena. London 21974, 66.
2C. Roberts, 'An Early Papyrus of the First Gospel', Harvard Theological Review 46 (1953), 233-7,
here 237, referring to the corroborative verdicts of H. Bell, T.C. Skeat and E.G. Turner.
3 C. Roberts, 'Complementary Note', in R. Roca-Puig, Un Papiro Griego del Evangelio de San Mateo,
Barcelona 21962, 59-60.
4 K. Aland's suggestion, based on an observation by P. Weigandt, in his 'Neue Neutestamentliche
Papyri II', New Testament Studies 12 (1965/66), here 193-5; repeated, as probable ("probablement") by J.
van Haelst, Catalogue des Papyrus littéraires juifs et chrétiens, Paris 1976, 146 (no. 403), and by C.H.
Roberts/T.C. Skeat, The Birth of the Codex, London 1983, 40-41.65-66, but discontinued by Aland
himself in his and his wife's Der Text des Neuen Testaments, Stuttgart 2 1989, 105, where ∏4 is
categorized as "Normaltext", but ∏64/67 as "fester Text", and 106/110, where the dates are given as "III"
and "um 200" respectively.
5 J. van Haelst, Catalogue des Papyrus littéraires juifs et chrétiens, Paris 1976, 125 (no. 336).
14 C.P. Thiede
of papyrus found in an envelope among the correspondence relating to the Matthean fragment, 6 a
mere 1.8 x 1.9 cm, with two fragmentary lines containing just two complete letters (Iota/Ny) and
five incomplete ones on one side (the other side is blank), in a later, larger script unrelated to the
three fragments of Matthew. The College Library now correctly numbers the three Matthean
fragments "Gr.17", and this should henceforth be the number used in all lists and catalogues of
NT papyri.
2) CONTENTS
There are some discrepancies between the editions of and references to Magdalen Gr.17:7
In his first edition of 1953, Roberts had transcribed the contents as Mt 26:7, 10, 14-15, 22-23, 31,
32-33.8 On two pages preceding his supplement to Roca-Puig's second edition of the Barcelona
fragments, he offered a new "Transcripción del P. Magd. de Oxford", translated into Spanish by
Roca-Puig.9
Dated "9.6. 60", this new transcription offers several alterations: Col II, recto (a), line 1 (Mt
26,31) now recognizes the nomen sacrum for Jesus as I% rather than IH; in line 2, the visible part
of the line is now extended to skandalisyh... rather than skanda...; and in Col II, recto (b), he
tentatively adds a new first line for verse 32, proaj[v and changes, in line 2, galeglaian to
galiglaian. Furthermore, he now corrects the contents; Mt 26:7-8, 10, 14-15, 22-23, 31, 32-
33.10 There is a very good reason for Roberts' insistence on the separation of v. 31 from vv. 32-
33: v. 31 is on a separate scrap of papyrus. The three scraps of Magdalen Gr.17, all of them with
text on recto and verso, offer six 'units', Mt 26: 7-8 (fr. 1, verso), 10 (fr. 2, verso), 14-15 (fr. 3,
verso), 22-23 (fr. 3, recto), 31 (fr. 1, recto), 32-33 (fr. 2, recto). Thus, Roberts' system with a
separate v. 31 is to be preferred for reasons of clarity and should be copied by Nestle-Aland et al.
The peculiar variants of Gr.17 were duly noted by Roberts; in three instances, however, he
himself seemed uncertain and mistaken, and one further variant has so far remained unnoticed.
i) fr. 3, verso, line 2 (26:14): Roberts had seen that d≈deka is written in the numerical
symbol ib – the lower half of the Beta is clearly visible. It is, however, equally obvious that there
is no space for an Omicron between Beta and the Lambda of legÒmenow. Thus, we have a rare
example of legÒmenow without the article, a construction paralleled by, e.g., Matthew 2:23 (efiw
pÒlin legom°nhn Nazar°t), John 4:5 (efiw pÒlin t∞w Samare¤aw legom°nhn Suxãr). In
Magdalen Gr.17/∏64, the omission may of course merely be a scribal error.
6I owe this information to K.S. Speirs, Assistant Librarian, Magdalen College Oxford, in a letter
dated 23rd February 1994.
7 Conversely, they all agree as to P.Barc.1 (∏67), Mt 3:9.15; 5:20-22.25-28
8 See note 2, here 236.
9 See note 3, here 57-59.
10 The 'standard' Greek New Testament, Nestle-Aland 271993, has 26: 7-8, 10, 14-15, 22-23, 31-33.
Papyrus Magdalen Greek 17 (Gregory-Aland ∏64). A Reappraisal 15
ii) fr. 3, recto, line 1 (26:22): Roberts had stated, in his 1953 edition,11 that "the papyrus
must have read legein eiw ekastow autv, an order which is unique ...". However, this unique
variant is far from evident. The three severely damaged letters in line 1 which Roberts had
identified as Tau/Omega/My are in fact Tau/Omega/Ny. The Ny is the final letter of autvn, and
thus our papyrus would have read … ßkastow aÈt«n mÆti §g≈ efimi …, the text of ∏ 45 , ∏37,
Bezae Cantab. (D), et mult.al. There is only one standard edition of the Greek NT which has this
as the best reading; 12 ∏ 64 now confirms the papyrological evidence for it against the text
preferred by other editions. 13
iii) fr. 1, recto, line 2 (Mt 26:31): For reasons of stichometry, the Íme›w after pãntew should
be omitted. With it, line 1 would have 2o rather than the average 16 letters. This omission, in the
oldest ms of Matthew's gospel, confirms, once again, the tendency of all early papyri to keep the
Greek simple, to pare it to the bones, free from rhetorical embellishments.14
iv) fr. 2, recto, line 2 (26:32): One has to read galeglaian a..., as Roberts had transcribed
it in his 1953 edition, not galiglaian a... as in the printing error of the second transcription
dated 1960. This variation of the common Galila¤an for Galilee is of course odd, and Roberts
adds a note to his 1960 edition which reads "v. 33, vel galeilaian.15 But apart from the fact that
it is v. 32, not 33, the Gamma is as unmistakable as the Epsilon which precedes it. Epsilon + Iota
for Iota is common enough to be unremarkable; as for Gamma instead of Iota, this is nothing but
a scribal error not quite inexplicable in view of the identity of the vertical strokes of Iota and
Gamma in this papyrus. The scribe of Magdalen Gr.17 was not averse to original decisions; even
this mistake is, in a way, original.
4) THE D ATE
The date commonly given to Magdalen Gr.17 (and P.Barc.1), ca. 200,16 may look like a
safe "dumping ground", but this might be too late. One has to keep in mind, of course, that
Roberts revolutionized the dating of the papyrus in his first edition of 1953, when he suggested
"a date in the later second century":17 At that time, he was confronted with the estimate provided
by the Rev. Charles B. Huleatt, a former demy (foundation scholar) of Magdalen College, who
had acquired the fragments at Luxor in 1901 and had given them to his old college in the same
year.18 Huleatt had suggested a date in the third century, and a note in the display cabinet with
Gr.17 in the Old Library of Magdalen College still reads: "2nd half of 3rd century (probable
date)". In the librarian's report of 1901, H.A. Wilson quoted an oral assessment from no less an
authority than A.S. Hunt who even thought that "they may be assigned with more probability to
the fourth century". As Roberts pointed out in his commentary,19 Hunt and his colleague B.P.
Grenfell had assumed, on principle, that manuscripts written in a codex could not be earlier than
the third, preferably the fourth century. He quotes the amusing example of P.Oxy. I,35, a Latin
codex fragment of an otherwise unknown History of the Macedonian Wars now at the British
Library,20 which they analyzed as belonging to the second century, perhaps even before AD 79 –
for palaeographical reasons –but which they nonetheless assigned to the late third or fourth
century because it is a vellum codex.21 As mentioned above, Roberts then proceeded to argue,
comparatively, for a late second century date of Magdalen Gr.17, backed in this by Bell, Skeat
and Turner. One of the decisive arguments he adduced is the fact that "in the Magdalen
fragments the minute omikron and the flat omega, common in third century hands, are absent".22
Since the publication of Roberts' paper, new papyri have become available, and they appear
to favour an even earlier date. This may not come as a surprise, since one tendency of the
reevaluation of NT papyri at least since the 60s has been a redating with, occasionally, somewhat
drastic and not undisputed consequences.23 It may be argued that the result of this continuing
process is a mounting degree of uncertainty, rather than certainty, as to the reliability of
palaeographical datings of literary hands; but even so, one should not eschew the challenge. For
Magdalen Gr.17/P. Barc.1., one such unexpected example is a leather scroll discovered in the
Nahal Hever, near the Dead Sea, the so called Greek Minor Prophets Scroll 8HevXIIgr.24 With
minor variations, D. Barthélemy (who first published parts of the scroll in 1963), C.H. Roberts,
W. Schubart, E. Würthwein and R. Hanhart all opt for c. AD 50/mid-first century AD.25 Tov, in
his new and complete edition, leaves the task of dating the scroll to P.J. Parsons who does not
rule out a mid- to late first century date by referring to P. Oxy. 2555, but prefers a date in the
later first century BC "as possible, though not of course necessary".26
Obviously then, there seems to be some scope for differing assessments, between the late
first century BC and the middle, or, at the latest, the second half, of the first century AD, with a
clear preference for the mid-first century AD. Without entering the debate about the date of
Matthew's gospel,27 we may note that the historical terminus post quem for any of the gospels
obviously is the year of the last events reported about the crucified and risen Jesus, AD 30, and
we may also note that this would give us enough space to accomodate a comparison between the
Nahal Hever Scroll and Magdalen Gr.17.
Even at first glance and using, as a point of reference, the plate in Schmidt/Thiel/Hanhart,
the identity and near-identity of several letters is striking: Alpha, Epsilon (a letter fluctuating in
both scripts), Iota, Omicron, Rho and Ny are particularly close. An equally obvious difference, on
20Brit.Lib. P. 745.
21The Oxyrhynchus Papyri Part I, London 1898, 59-60, here 59.
22 See note 2, here 237.
23 E.g., H. Hunger, 'Zur Datierung des Bodmer II', Anzeiger der phil.-hist. Klasse der Österr. Akad.
der Wissenschaften, 4 (1960), 12-23, arguing for a date in the first quarter of the second century, against
the traditional consensus which dates Bodmer II (∏66) to "c. 200"; or Y.K. Kim, 'Palaeographical Dating
of ∏46 to the Later First Century', Biblica 69 (1988), 248-57, taking away some one-hundred years from
the usual dating of P. Chester Beatty II/University of Michigan Inv. 6238, "c. 200". See also note 32
below.
24 E. Tov, ed., The Greek Minor Prophets Scroll 8HevXIIgr, Oxford 1990.
25 For Barthélemy, Roberts, Schubart and Würthwein see the summary in E. Würthwein, Der Text des
Alten Testaments, Stuttgart 51988, 194, with plate; for Hanhart, see W.H. Schmidt/W. Thiel/R. Hanhart,
Altes Testament, Stuttgart et.al. 1989, 194-5, with plate.
26 P.J. Parsons, 'The Scripts and Their Date', in E. Tov, as in note 15, 19-26, here 24 and 26.
27 For a survey, see D. Guthrie, New Testament Introduction, Leicester 41990, 43-57.
Papyrus Magdalen Greek 17 (Gregory-Aland ∏64). A Reappraisal 17
the other hand, may be seen in the Etas and Mys; but the second scribe of the Nahal Hever scroll
provides the comparable Eta and My more than once.28 The Nahal Hever scroll of the Minor
Prophets may be at the extreme end of the spectrum, but is not the only first century analogy.
Further material is provided by papyri in the script of Herculaneum, for which AD 79 is the
natural focal point. 29 Interestingly, there is a small, unidentified Greek fragment from Qumran
Cave 7, 7Q61 , for which the archaeological terminus ante quem is AD 68, which has the
characteristic Eta with the horizontal stroke above the median, evident in Magdalen Gr. 17.30
There also is a Greek papyrus from Qumran Cave 4 which shows several letters resembling
Papyrus Magdalen Gr.17, such as the Alpha, the Beta, etc.: pap4QLXXLeviticusb. As Parsons
points out, the script is far from uniform, but this papyrus from Cave 4 could be dated to the mid-
first century AD.31
Unwittingly, he then procedes to offer an interesting case study: In his drawings of letters
of the preceding fragment 4QLXXLeva (parts of a leather scroll which he dates to the first
century BC), the Alpha, Beta, Delta, Epsilon, Eta, Iota, Kappa, Eta etc are identical or near-
identical to what we find in Magdalen Gr.17. In fact, the letters he draws could have been taken
straight out of Gr.17. Looking at the fragments themselves, there would seem to be at least two
differences, however: the Qumran Leviticusa is sloping slightly to the right, and the letters are
very close to each other, occasionally even connected (ligatures). Even so, archaeology alone
cannot have influenced Parsons' very early dates – there is scope until AD 68, after all, when the
caves were abandoned, and one might well prefer mid-first century dates AD for both
4QLXXLeva and pap4QLXXLevb . But, and this is the point, the prevailing tendency to date
material of a nature comparable to Magdalen Gr.17 to a period even preceding the earliest
possible date of Matthew's gospel suggests, with all due caution, the possibility of redating the
fragments from Oxford and Barcelona – which are, after all, definitely Matthean – to a period
somewhat earlier than the late second century previously assigned to them. Certainty will remain
elusive, of course.
To sum up, even though Herculaneum and Qumran (with its Greek fragments in two caves,
4 and 7) are still under survey, they both have their archaeological termini: all comparative
material taken directly from their finds suggest dates prior to AD 79 and 68 respectively. It goes
without saying that scribal characteristics found in those places may well have continued to be in
use afterwards, towards the end of the first century, and occasionally even later. For our present
purposes, we may proffer a tentative suggestion: the material from Nahal Hever, Herculaneum
and Qumran could point towards a first century date for Magdalen Gr.17 / P.Barc.1.
At this stage, we must turn to the Nomina Sacra and their influence on the date: Magdalen
Gr.17 has two, probably three abbreviations of holy names and words: iw for ÉIhsoËw (fr. 2,
verso, l. 1, probable because of the stichometry of the line; fr. 1, recto, l. 2, definitely) and ke for
28 For a single plate, see E. Würthwein, as in note 16, 195; for the two scribes, see P.J. Parsons, as in
note 17, here 20.
29 See W. Schubart, Griechische Paläographie, München 21966, 111; J. O'Callaghan, 'Paleografia
Herculanense en algunos papiros griegos de Qumrân', in Homenaje a Juan Prado, Madrid 1975, 529-32.
30 M. Baillet / J.T. Milik / R. de Vaux, eds., Les 'Petites Grottes' de Qumrân, DJD III, Oxford 1962,
companion vol. 'Planches', plate XXX.
31 P.J. Parsons, The Palaeography and Date of the Greek Manuscripts, in P.W. Skehan/E. Ulrich/J.E.
Sanderson, eds., Qumran Cave 4, IV, DJD IX, Oxford 1992, 7-13, here 8. Such references to Greek ms
from Qumran are all the more legitimate as they did not originate at Qumran and could represent a wide
variety of Jewish diaspora hands. Cf. E. Tov, 'Hebrew Biblical Manuscripts from the Judaean Desert:
Their Contribution to Textual Criticism', Journal of Jewish Studies 39/1 (1988), 5-37, here 19; et al.
18 C.P. Thiede
kÊrie (fr. 3, recto, l.2). For historical reasons, Roberts had suggested that the use of these and
other nomina sacra had become established practice among Christians in Jerusalem even before
the year AD 70.32 He did lack the palaeographical evidence, though, and even for John Rylands
Gr. 457 (∏ 52 ), which he himself had edited,3 3 he did not suggest nomina sacra in the
reconstructed, missing parts of the extant lines, although this would have been possible within the
given stichometry.34 Magdalen Gr.17 might offer the missing link: a Christian codex fragment of
the first century, perhaps (though not necessarily) predating AD 70, with the nomina sacra
postulated by Roberts.35
Some time ago, such a date would have been ruled out for the simple reason that a copy of
a codex of Matthew – and there is no dispute whatsoever about the identification of the Oxford
and Barcelona papyri – cannot have reached Egypt at such an early stage of the gospel's
germination and transmission. But we have learned from the methodological error of Grenfell
and Hunt, described above; and we possess that famous fragment of a codex of John's gospel kept
at the John Rylands University Library Manchester, ∏52 (J. Rylands Libr. Gr.P. 457). Within the
range of dated and datable papyri Colin Roberts compared to ∏ 52 for his first edition in 1935, he
finally decided in favour of what is arguably the latest possible date, c. AD 125. He could,
however, have been less cautious by preferring the other end of the spectrum, documented by P.
Fayyum 110 of AD 94 or P. Lond. 2078, a private letter from the time of Domitian, AD 81-96.
There also is a good resemblance to ∏52 in P.Gr. Berol. 19c, part of a scroll with Iliad X, from the
end of the 1st century.36
In their monograph, The Birth of the Codex, C.H. Roberts and T.C. Skeat argue that the
Christians had chosen the codex form for copies of Old Testament texts and their own writings
before AD 100.37 Near the end of the first century, the Roman poet Martial praises his and his
publisher friend Secundus' unheard-of marketing enterprise, the introduction of a library of
classical works in the codex format; the Latin codex fragment of a History of the Macedonian
32 C.H. Roberts, Nomina sacra: Origins and Significance, in id., Manuscript, Society and Belief in
Early Christian Egypt, London 1969, 26-48, here 46. Cf also J. O'Callaghan, "Nomina sacra" in papyris
graecis saeculi III neotestamentariis, Rom 1970, for an analysis of nomina sacra in ∏ 46 which may,
according to Kim, be late first century. B.M. Metzger, The Text of the New Testament. Its Transmission,
Corruption and Restoration, 3rd, enlarged ed., Oxford 1992, 265-6, thinks that it is precisely the
occurrence of nomina sacra in ∏ 46 which counts against Kim's advocacy of such an early date. But
whatever the quality of Kim's arguments as such, this could well be a circular argument on Metzger's side.
If, for other palaeographical reasons, certain papyri with nomina sacra turn out to be first century, then
this would favour Roberts' theory of the early origins of nomina sacra rather than Metzger's preference for
later dates.
33 See C.H. Roberts, An Unpublished Fragment of the Fourth Gospel in the John Rylands Library,
Manchester 1935; enlarged and amended in Bulletin of the John Rylands Library 20 (1936), 45-55.
34 See C.P. Thiede, The textual peculiarities of ∏ 52, in id., The Earliest Gospel Manuscript ?, Exeter
1992, 13-19, here n. 9, 17-18.
35 It seems to me that the "watershed" is the Christian change from scroll to codex, most likely some
time after the destruction of the temple in the year AD 70 which contributed to the end of Jewish-
Christian missionary activities among their fellow Jews and terminated the strategical reasons for using
the scroll format and for resisting the temptation to put Jesus on a par with God (Jahwe) palaeographically
by means of nomina sacra which had, until then, been the privilege of Jewish scribes using the
tetragrammaton for the name of God.
36 See W. Schubart, as in note 29, 117-8.
37 The Birth of the Codex, London 1983, 61.63. Cf, more recently, T.C. Skeat, 'The Origin of the
Christian Codex', ZPE 102 (1994), 263-268.
Papyrus Magdalen Greek 17 (Gregory-Aland ∏64). A Reappraisal 19
Wars mentioned above may be the only surviving example of this possibly short-lived venture.38
Under the influence of Roberts' cautious dating of ∏ 52 and five years before Kim's paper,39
Roberts and Skeat do not name any first century Christian codex to corroborate their theory with
some practical evidence. The present state of affairs, however, suggests that the Oxford
fragments Magdalen Gr.17, with their Spanish counterparts, would be among the prime examples
of the birth of the Christian codex prior to the turn of the century.
The fragments of Matthew's gospel in the Old Library of Magdalen College Oxford,
henceforth to be listed as Magdalen Greek 17 rather than 18, remain the oldest extant papyrus of
that gospel; but it may be argued that it could be redated from the late second to the late first
century, some time after the destruction of the Temple in Jerusalem. It appears to be the oldest
known codex with nomina sacra. Lists of New Testament papyri should reflect the fact that the
three fragments of Magdalen Gr.17 = ∏64 preserve text on all six sides, not just on five, as is the
impression conveyed at present.
∏64 is a "new", additional witness to a construction of Mt 26:22 preferred by the papyri ∏37
and ∏45 and several later ms, but ruled out by the most widely used editions of the Greek New
Nestament, Nestle-Aland 27(1993) and The Greek New Testament UBS4 (1993). The accumulated
evidence now clearly suggests ... ßkastow aÈt«n mÆti ... as the better text, and this should be
acknowledged by future editions of the Greek NT, in concurrence with Bover-O'Callaghan
3 1994. This improved reading and two further variants, legÒmenow without the article in 26:14
and, in particular, the likely omission of Íme›w after pãntew in 26:31, appear to confirm the
impression that the very earliest papyri tend to preserve a simple but clear and effective Greek
untouched by the literary ambitions of later scribes.
38
See note 20 and Martial, Epigrams, I,2; see also Roberts/ Skeat, as in note 24, 24-29, E.G. Turner,
The Typology of the Early Codex, Pennsylvania 1977, and J. van Haelst, Les Origines du Codex, in A.
Blanchard, ed., Les Débuts du Codex, Turnhout 1989, 13-35.
39 See note 23.
20 C.P. Thiede