Hostname: page-component-5d59c44645-zlj4b Total loading time: 0 Render date: 2024-02-24T01:42:55.365Z Has data issue: false hasContentIssue false

The archetype of Beowulf

Published online by Cambridge University Press:  26 September 2008

Michael Lapidge
University of Notre Dame


It is a cardinal tenet of classical textual criticism that medieval scribes were most prone to error when copying from an unfamiliar system of script. Accordingly a good deal of attention has been given by classical scholars to what happens to a text when it is copied from one system of script to another, and to the characteristic sorts of error which such copying involves. The great French textual critic, Alphonse Dain, even coined a Greek term, metacharakterismos (μεταχαρακτηρισμός), to describe the scribal process of copying, character by character, from one script to anodier. (The Latin equivalent would be translitteratio, which might be rendered ‘transliteration’ in English.) Dain was thinking principally of the transliteration of Greek uncial manuscripts into minuscule script; but the process is also known to have taken place in the transmissional histories of Latin texts, when works of classical literature in (say) rustic capital script were transliterated into the various regional minuscules. By observing patterns of repeated error, Latin textual critics have often been able to demonstrate that the archetype of such-and-such a text must have been written in a particular system of script. The first attempt at such a demonstration was apparently that by the humanist scholar Joseph Scaliger (1540–1609), who in his Castigationes in Catullum (1577) showed that the archetype of all surviving manuscripts of Catullus was written in what he called Langobardicae litterae, what we should describe as a form of pre-Caroline minuscule.

Research Article
Copyright © Cambridge University Press 2000

Access options

Get access to the full version of this content by using one of the access options below. (Log in options will check for institutional or personal access. Content may require purchase if you do not have access.)


1 Dain, A., Les manuscrits, 3rd ed. (Paris, 1975), pp. 124–35.Google Scholar

2 See in particular the incisive comments of Timpanaro, S., La genesi del metodo di Lachmann, 3rd ed. (Padua, 1985), pp. 111–21Google Scholar (‘Sulla determinazione del tipo di scrittura di codici perduti’).

3 On Scaliger, see briefly von Wilamowitz, U., History of Classical Scholarship, ed. Lloyd-Jones, H. (London, 1982), pp. 4953, and esp.Google ScholarGrafton, A., Joseph Scaliger: a Study in the History of Classical Scholarship, 2 vols. (Oxford, 19831993).Google Scholar Vol. I is concerned with Scaliger's contribution to textual criticism.

4 Timpanaro, , La genesi, p. 10Google Scholar, and Grafton, A. T., ‘Joseph Scaliger's Edition of Catullus (1577) and the Traditions of Textual Criticism in the Renaissance’, Jnl of the Warburg and Courtauld Institutes 38 (1975), 155–81, at 170–3.CrossRefGoogle Scholar On what was meant in the Renaissance by Langobardicae litterae, see Rizzo, S., Il lessico filologica degli umanisti (Rome, 1973), pp. 122–3.Google Scholar

5 T. Lucreti Cari De Rerum Natura, ed. Lachmann, K., 2 vols. (Berlin, 1850).Google Scholar

6 See Timpanaro, , La genesi, pp. 63–8Google Scholar and passim.

7 De Rerum Natura, ed. Lachmann, II, 3Google Scholar: ‘id exemplar ceterorum ARCHETYPON (ita appellate soleo) constitit paginis CCCII, quarum non tantum prima et ultima, sed praeterea centesima nonagesima, quae erat post finem libri quarti, conscriptae non fuerunt … porro eius codicis scripturam litteris capitalibus gracilioribus, non uncialibus, fuisse multis indiciis cognosci potest: vocabulorum distinctio nulla fuit, sententiarum constantissima in mediis versibus. ex quibus rebus intel-legitur eum saeculo post Christum natum quarto quintove conscriptum fuisse’. The pagination and layout of the text of the reconstructible archetype have been set out in tabular form by Goold, G. P., ‘A Lost Manuscript of Lucretius’, Acta Classica 1 (1958), 2130.Google Scholar See also Timpanaro, , La genesi, pp. 113–19Google Scholar; and, briefly, Texts and Transmission: a Survey of the Latin Classics, ed. Reynolds, L. D. (Oxford, 1983), pp. 218–22.Google Scholar

8 Duvau, L., ‘Lucretiana’, Revue de philologie, de littérature et d'histoire anciennes 12 (1888), 31–7, esp. 35–6Google Scholar, where a series of literal confusions attested in the transmission of Lucretius (notably between a and u, r and n, and c and t) suggested to Duvau that the hyparchetype was written in a form of Insular minuscule.

9 Ribbeck, O., Prolegomena Critica ad P. Vergili Maronis Opera Maiora (Leipzig, 1866), pp. 231–64 and 381454Google Scholar; see also Gaebel, R. E., ‘Roman Cursive Influence in the Text of the Georgics’, Rheinisches Museum 128 (1985), 305–13.Google Scholar

10 See Texts and Transmission, ed. Reynolds, passim, as well as Brunhölzl, F., ‘Zu den sogennanten “codices archetypi” der römischen Literatur’, in Festschrift Bernhard Bischoff zu seinem 65. Geburtstag, ed. Autenrieth, J. and Brunhölzl, F. (Stuttgart, 1971), pp. 1631Google Scholar, and Zelzer, M., ‘Die Umschrift lateinischer Texte am Ende der Antike und ihre Bedeutung für die Textkritik’, Wiener Studien 94 (1981), 211–31.Google Scholar There is a valuable survey of palaeographical evidence for Insular stages in the prehistories of classical Latin transmissional histories by Dumville, D. N., ‘The Early Mediaeval Insular Churches and the Preservation of Roman Literature: towards a Historical and Palaeographical Reëvaluation’, in Formative Stages of Classical Traditions: Latin Texts from Antiquity to the Renaissance, ed. Pecere, O. and Reeve, M. D. (Spoleto, 1995), pp. 197237.Google Scholar

11 There are valuable (but random rather than systematic) observations by Gerritsen, J., ‘Have with you to Lexington! The Beowulf Manuscript and Beowulf’, in In Other Words. Transcultural Studies in Philology, Translation and Lexicography presented to Prof. dr. H. M. Meier on the Occasion of his Sixty-Fifth Birthday (Dordrecht, 1989), pp. 1534, at 24Google Scholar (confusion of a and u), and Clemoes, P., Interactions of Thought and Language in Old English Poetry, CSASE 12 (Cambridge, 1995), 33–4 (confusion of d and ð and a and u).CrossRefGoogle Scholar

12 The exception to this generalization is Kiernan, K. S., Beowulf and the Beowulf Manuscript, 2nd ed. (Ann Arbor, MI, 1996)Google Scholar, who argues that the manuscript is autograph, and that Scribe II is the poet himself who artistically combined two originally distinct Beowulf narratives (ibid. pp. 243–78, esp. 271). Kiernan's arguments have been challenged from many quarters: see Gerritsen, J., ‘Beowulf Revisited’, ES 79 (1998), 82–6, esp. 82, n. 2.Google Scholar

13 The most concerted attempts have been by Sisam, K.: Studies in the History of Old English Literature (Oxford, 1953), pp. 6596Google Scholar (concerned with the prehistories of the texts which accompany Beowulf in the Nowell Codex, rather than with the poem itself), and The Structure of Beowulf (Oxford, 1965), pp. 6771 (The Transmission of Beowulf).Google Scholar

14 See the concise discussion by Andrew, S. O., Postscript on Beowulf (Cambridge, 1948), p. 133.Google Scholar

15 The manuscript is described by Ker, N. R., Catalogue of Manuscripts containing Anglo-Saxon (Oxford, 1957), pp. 281–3Google Scholar (no. 216). The entire manuscript is ed. in facsimile by Malone, K., The Nowell Codex, EEMF 12 (Copenhagen, 1963).Google ScholarBeowulf on its own is ed. in facsimile by Zupitza, J., Beowulf, 2nd ed. rev. N. Davis, EETS os 245 (Oxford, 1959)Google Scholar; and see now The Electronic Beowulf, ed. Kiernan, K., 2 CD-ROMs (London, 1999).Google Scholar

16 The Passio S. Christophori is ed. Rypins, S., Three Old English Prose Texts in MS. Cotton Vitellius A. xv, EETS os 161 (Oxford, 1924), 6876Google Scholar; for The Wonders of the East and The Letter of Alexander to Aristotle, one uses the editions in Orchard, A., Pride and Prodigies. Studies in the Monsters of the Beowulf-Manuscript (Cambridge, 1995), pp. 184203Google Scholar [The Wonders of the East] and 224–53 [The Letter of Alexander to Aristotle]. For Judith, see the recent edition by M. S. Griffith (Exeter, 1997). For editions of Beowulf, see below, pp. 9–10.

17 Because the last page of Beowulf (209v) is badly abraded, it has been thought that it originally formed the conclusion of the volume. But although Judith was unquestionably copied by Scribe II, various scholars have suggested that it originally preceded the Passio S. Christophori, and was relocated when the manuscript was rebound. See, for example, Lucas, P. J., ‘The Place of Judith in the Beowulf–Manuscript’, RES 41 (1990), 463–78.Google Scholar

18 The crucial discussion of the date is Dumville, D. N., ‘Beowulf Come Lately. Some Notes on the Palaeography of the Nowell Codex’, ASNSL 225 (1988), 4963Google Scholar (repr. in his Britons and Anglo-Saxons in the Early Middle Ages (Aldershot, 1993), no. VII); cf. also idem, The Beowulf-Manuscript and How Not to Date It’, Med. Eng. Stud. Newsletter 39 (1998), 21–7Google Scholar (replying to arguments incorporated by Kiernan, K. S. in the second edition of his Beowulf and the Beowulf Manuscript, pp. xvxxx).Google Scholar

19 Andrew, (Postscript on Beowulf, p. 133Google Scholar) notes that, in Klaeber's edition of the poem, some 250 corrections have been admitted. Cf. the remark of Tolkien, J. R. R., The Old English Exodus, ed. Turville-Petre, J. (Oxford, 1981), p. 34Google Scholar: ‘The Beowulf manuscript… is as free from these complexities as it is from such costly elaboration. If the text of Beowulf is full of the minor accidental errors, it is as a whole a remarkable record, whose general fidelity, even when its scribes clearly were at sea, bears close examination.’

20 See the apparatus criticus in Orchard, Pride and Prodigies: Wonders of the East, c. 5 n. a–a; cc. 33–7 are totally omitted by Scribe I. To judge from the layout of the text, however, it is probable that the omission of cc. 33–7 is due to physical loss rather than scribal oversight: the present second quire (fols. 104–9) is a quire of six, and the simplest explanation is that a bifolium containing cc. 33–7 has fallen out from the middle of the quire, from between what are now fols. 106–7.

21 For example the word leuua (‘league’), which he invariably copied as the meaningless leones (half a dozen times).

22 On the accuracy (or otherwise) of scribal performance in Beowulf, see Andrew, , Postscript on Beowulf, pp. 133–52Google Scholar (‘Scribal Error and its Sources’).

23 See lines 139 (sobte), 149 (secgum), 586 (swipe), 1130 (ne), 1174 (pe), 1329 (æ peling), 1372 (hydan), 1404 (pær), 1546 (ond), 1559 (wæs) and 1889 (heap). Half-lines are omitted in lines 403 and 1803, and a line has been omitted in 389–90.

24 Cf. Hulbert, J. R., ‘The Accuracy of the B-Scribe of Beowulf’, PMLA 43 (1928), 1196–9.CrossRefGoogle Scholar

25 See lines 2139 (grund), 2251 (lif), 2488 (hearo), 2589 (ofer), 2620 (pa), 2694 (gefrægn), 2941 (fuglum), 3000 (wen), 3086 (peodcyning) and 3101 (siðe). A half-line is omitted from line 2792.

26 Cf. Andrew, , Postscript on Beowulf, p. 133Google Scholar: ‘there are over sixty words in which a letter is dropped or inserted or changed’.

27 The reference to ‘Grundtvig (1820)’ is not properly to an edition - his edition of the poem only appeared in 1861 - but to the textual notes in his Bjowulfs Drape: Et Gothisk Helte-Digt fra forrige Aar-Tusinde af Angel-Saxisk paa Danske Riim (Copenhagen, 1820), pp. 267312.Google Scholar

28 Kelly, B., ‘The Formative Stages of Beowulf Textual Scholarship’, ASE 11 (1983), 247–74Google Scholar [Part I], and 12 (1984), 239–75 [Part II‘Kelly I’ and ‘Kelly II’, respectively. For bibliographical details of the specified editions, see Greenfield, S. B. and Robinson, F. C., A Bibliography of Publications on Old English Literature to the End of 1972 (Manchester, 1980)Google Scholar, nos. 1632, 1659, 1633, 289, 1635, 1637 and 1638 respectively.

29 In what follows I often state that a correction has been accepted by ‘all editors to 1983’, by which I usually mean that it has been tabulated in Kelly II as ‘all’; but note that, for no obvious reason, Kelly did not include any of the three editions of the poem by W J. Sedgefield, and that her attributions of corrections are occasionally inaccurate. But these minor imperfections do not mar an exceptionally useful article.

30 Although I am discussing the transmission of an Old English text, my palaeographical examples will perforce be drawn from Latin manuscripts. However, it will be remembered that the same script was used for writing both Latin and English in the period before the late tenth century, when, after the introduction of Caroline minuscule, scribes began to distinguish between Latin text (for which they used Caroline script) and vernacular text (Square minuscule, and its successor in the eleventh century, English vernacular minuscule).

31 Bugge, S., ‘Zum Beowulf’, ZDA 4 (1873), 192224, at 197Google Scholar, who argued that the prefix un- in the transmitted unhar was an intensifier.

32 Cf. also Dream of the Rood 117 anforht (MS unforht).

33 Trautmann, M., ‘Berichtigungen, Vermutungen und Erklärungen zum Beowulf. Erste Hälfte’, Banner Beiträge zur Anglistik 2 (1899), 121–92Google Scholar, at 147 (and adopted in his edition of 1904).

34 Cf. also Andreas 1545 wadu (MS wudu).

35 Barnouw, A. J., Textkritische Untersuchungen nach dem Gebrauch des bestimmten Artikels und des schwachen Adjectivs in der altenglischen Poesie (Leiden, 1902), p. 36.Google Scholar

36 The number of such cases would be increased further if one were to admit the four cases where the transmitted text has sunu, but where one or more editors would emend to suna. (For the ‘normal’ declension of sunu, see Campbell, A., Old English Grammar (Oxford, 1959), § 611.Google Scholar) Thus in 344 (‘wille ic asecgan sunu Healfdeanes’), where the dative is required, Kemble and other editors emend to suna; in 1115 (‘het ða Hildeburh æt Hnæfes ade / hire selfre sunu sweoloðe befætan’), where Thorkelin and Grein emend to suna, even though the transmitted form sunu appears to supply the requisite accusative; in 1278 (‘sorhfulne sið sunu deoð wrecan’), where Grendel's mother sets off to avenge her son's death, and where some editors have emended to suna so as to supply the usual form of the genitive; and 1808 (‘heht Þa. se hearda Hrunting beran / sunu Ecglafes heht his sweord niman’), where Grundtvig emended to suna on the assumption that the dative is here intended, but where other editors have expressed doubt (see Klaeber's note ad loc). But because there is confusion between -u and -a in the nominative and oblique cases of the u–declension from an early date onwards (cf. Campbell, , Old English Grammar, § 613Google Scholar), most editors retain the transmitted forms rather than emending to suna. Because of this prevailing uncertainty, I have not listed the four instances of sunu among the examples of a and u confusion, nor are they listed by Kelly, ‘The Formative Stages’.

37 The essential studies of Anglo-Saxon Square minuscule are Dumville, D. N., ‘English Square Minuscule Script: the Background and Earliest Phases’, ASE 16 (1987), 147–79Google Scholar, and idem, English Square Minuscule Script: the Mid-Century Phases’, ASE 23 (1994), 133–64.Google Scholar There are also two brief studies of individual examples of the script: Roper, M., ‘A Fragment of Bede's De temporum ratione in the Public Record Office’, ASE 12 (1983), 125–8Google Scholar, and Ganz, D., ‘An Anglo-Saxon Fragment of Alcuin's Letters in the Newberry Library, Chicago’, ASE 22 (1993), 167–77.Google Scholar

38 Listed H. Gneuss, Handlist of Anglo-Saxon Manuscripts (forthcoming), no. 174 (note that, except in cases of supplementary items, the numbers in the Handlist are the same as those in Gneuss, H., ‘A Preliminary List of Manuscripts Written or Owned in England up to 1100’, ASE 9 (1981), 160).Google Scholar The manuscript is discussed and illustrated in Keynes, S., Anglo-Saxon Manuscripts and other Items of Related Interest in the Library of Trinity College, Cambridge, OEN Subsidia 18 (Binghamton, NY, 1992), 1617 and pl. VI.Google Scholar

39 Listed Gneuss, Handlist, no. 257; ed. in facsimile as The Exeter Book of Old English Poetry, ed. Chambers, R. W., Förster, M. and Flower, R. (Bradford, 1933).Google Scholar The illustration is from Guthlac B, 890–2.

40 Examples and brief descriptions of the various grades of script discussed here – Insular Half-uncial, hybrid, set and cursive minuscule – may conveniendy be found in Brown, M. P., A Guide to Western Historical Scripts from Antiquity to 1600 (London, 1990), nos. 16–19.Google Scholar For the terminology used to describe these script types, see the following note.

41 The terminology now used by palaeographers to describe the various grades and types of Insular script is that of the late Julian Brown, as expounded in his legendary (but unfortunately unpublished) Lyell Lectures delivered to the University of Oxford in 1977. In several of his publications he alluded briefly to his system of terminology: A Palaeographer's View: the Selected Writings of Julian Brown, ed. Bately, J., Brown, M. P. and Roberts, J. (London, 1993), pp. 179220passim.Google Scholar There is a searching critique of Brown's classification (and especially his dating) of Insular script by Dumville, D. N., A Palaeographer's Review: the Insular System of Scripts in the Early Middle Ages I (Kansai, 1999).Google Scholar

42 A Palaeographer's View, ed. Bately, et al. , p. 201Google Scholar: ‘the very high grade of minuscule which I call “hybrid”, since it borrows from half-uncial the general roundness of aspect that is natural to a straight-pen but not to a slanted-pen script, and uses, exclusively or as alternatives, some forms of letters which are characteristic of half-uncial rather than minuscule: the o-c form of a, which Lowe regarded as the “shibboleth” for half-uncial, and uncial d, n, r, and s’.

43 Listed Gneuss, Handlist, no. 432, and Lowe, E. A., Codices Latini Antiquiores, 11 vols. and suppl. (Oxford, 19341971; 2nd ed. of vol. II, 1972)Google Scholar [hereafter cited as CLA] II, no. 199.

44 A Palaeographer's View, ed. Bately, et al. , p. 201Google Scholar: ‘“Set” minuscule on the other hand obtains greater formality by lifting the pen at all the points at which cursive minuscule normally links the parts of letters.”

45 Listed Gneuss, Handlist, no. 450, and CLA II, no. 215.

46 Listed Gneuss, Handlist, no. 604, and CLA II, no. 235.

47 Listed Gneuss, Handlist, no. 635, and CLA II, no. 241.

48 A Palaeographer's View, ed. Bately, et al. , p. 201Google Scholar: ‘The basic grade was “cursive” minuscule, a narrow, pointed script for which a pen of moderate thickness was cut at right angles to the shaft, held at the natural “slanting” angle, and used to write an alphabet in which a number of letters were formed without or nearly without penlifts, notably a, h, m, n and f, p, r, s. An accelerated version of cursive minuscule, which may be called “current”, sometimes links letters as well as parts of letters and makes full use of the numerous ligatures, and abbreviations that were known to all Insular scribes.’

49 Listed Gneuss, Handlist, no. 846, and CLA XI, no. 1618; illustrated in Facsimiles of the Creed, ed. Burn, A. E., HBS 36 (London, 1909), pl. XVIII.Google Scholar

50 Parkes, M. B., ‘The Handwriting of St Boniface: a Reassessment of the Problems’, BGDSL 98 (1976), 161–79Google Scholar, repr. in his Scribes, Scripts and Readers. Studies in the Communication, Presentation and Dissemination of Medieval Texts (London, 1991), pp. 121–42.Google Scholar

51 Listed Gneuss, Handlist, no. 846, and CLA XI, no. 1621; ed. in facsimile by Arngart, O., The Leningrad Bede. An Eighth-Century Manuscript of the Venerable Bede's Historia Ecclesiastics Gentis Anglorum in the Public Library, Leningrad, EEMF 2 (Copenhagen, 1952).Google Scholar For the identification of the four scribes, see ibid. pp. 18–19, and discussion by Parkes, , Scribes, Scripts and Readers, pp. 100–6.Google Scholar

52 Listed Gneuss, Handlist, no. 154, and CLA Suppl., no. 1679.

53 Listed Gneuss, Handlist, no. 824, and CLA VI, no. 760. The manuscript is ed. in facsimile by Sweet, H., The Epinal Glossary (London, 1883)Google Scholar, and more recently by Bischoff, B., Budny, M., Harlow, G., Parkes, M. B. and Pheifer, J. D., The Epinal, Erfurt, Werden, and Corpus Glossaries, EEMF 22 (Copenhagen, 1988).Google Scholar The early date of the manuscript (which had previously been dated as late as the ninth century by Sweet) was first mooted by Brown, T. J. (A Palaeographer's View, ed. Bately, et al. , p. 286, n. 12Google Scholar: ‘my suspicion that Epinal was copied in Mercia, saec. VII–VIII’), and argued more fully by Bischoff, and Parkes, , ‘Palaeographical Commentary’, in The Epinal, Erfurt, Werden and Corpus Glossaries, pp. 1317, esp. 16.Google Scholar The attribution to Malmesbury is my own conjecture, based on consideration of the intimate links between the ‘Epinal Glossary’ itself and the writings of Aldhelm - so intimate, in fact, that it is often impossible to tell which is the source of the other (cf. Pheifer, J. D., Old English Glosses in the Epinal-Erfurt Glossary (Oxford, 1974), pp. lvi and xci)Google Scholar – and the fact that the glossary, though written in a high grade of script, shows frequent signs of correction and alteration in a lower grade, as if it was being worked over by the compiler himself, rather than by a copyist. In any case there is general agreement that the manuscript was written in Southumbria c. 700.

54 For example, the continental scribes who copied from the Fulda manuscript of Bede (their copies are preserved in St Gallen and Munich) very frequently rendered the open a as u: see Lapidge, M., ‘Prolegomena to an Edition of Bede's Metrical Vita S. Cuthberti’, Filologia Mediolatina 2 (1995), 127–63, at 149–50.Google Scholar

55 There are numerous examples of eighth-century Anglo-Saxon scribes making frequent, but not consistent, use of open a while writing set minuscule script: CLA II, nos. 196b (an Anglo-Saxon scribe at Tours, s. viii2), 216 (s. viii/ix), Suppl. nos. 1687 (s. viii2) and 1792 (the Miskolc fragment of Aldhelm, s. viii1, in which open a and u are virtually indistinguishable: note the word nascuntur in Lowe's plate). The open a is also found in a manuscript of mid-ninth-century date (CLA II, no. 234). The scribe of this manuscript also wrote two original charters of King Æthelwulf, dated 839 and 847 (S 287 and 298). Peter Clemoes, basing himself on the opinion of David Dumville, noted that the latest dated use of documentary open a is found in this charter of 847 (S 298): Interactions, p. 33, n. 77. Later uses in literary manuscripts are exceptionally rare: see below, n. 86.

56 In spite of my earlier undertaking not to cite examples from illegible pages of the manuscript, notably 179r and 198v, I have included this example, because the word wonn is easily legible on 198v, even though words preceding and following it are not so.

57 Bugge, S., ‘Studien über das Beowulfepos’, BGDSL 12 (1887), 1112 and 360–75, at 111.Google Scholar

58 In 1141 (‘gif he torngemot þurhteon mihte / þæt he Eotena beam inne gemunde’), Trautmann, followed by Klaeber, emended the transmitted inne to irne, a syncopated form of the adjective irenne. But this emendation has not been adopted by subsequent editors, and has been questioned on metrical grounds by Fulk, R. D., ‘Textual Criticism’, in A Beowulf Handbook, ed. Bjork, R. E. and Miles, J. D. (Lincoln, NE, 1997), pp. 3553, at 42Google Scholar: ‘Even Klaeber makes some metrically poor judgments, as when he adopts Trautmann's (1904) emendation of manuscript “inne” to “irne” (1141 b) requiring trisyllabic scansion, as it would stand metrically for irenne

59 Listed Gneuss, Handlist, no. 343, and CLA II, no. 187. The manuscript is ed. in facsimile by T. D. Kendrick, T. J. Brown, R. L. S. Bruce-Mitford, H. Roosen-Runge, A. S. C. Ross, E. G. Stanley and A. E. A. Warner, 2 vols. in 3 pts (Olten, 1956–1960).

60 Listed Gneuss, Handlist, no. 45, and CLA II, no. 122; the manuscript is ed. in facsimile by Bischoff et al., The Epinal, Erfurt, Werden and Corpus Glossaries.

61 Listed Gneuss, Handlist, no. 910, and CLA I, no. 87. See Brown, , A Palaeographer's View, pp. 191, 212 and 215Google Scholar, but note also the reservations of Dumville, , A Palaeographer's Review, p. 73 and n. 78.Google Scholar The manuscript is ed. in facsimile by Brown, T. J. and Mackay, T. W., Codex Vaticanus Palatinus Latinm 235: an Early Insular Manuscript of Paulinus of Nola, Carmina (Turnhout, 1988).Google Scholar

62 There are numerous examples of the confusion of r and n in the corpus of Old English poetry: Genesis A 2252 Agar (MS agan), 2645 pæne (MS pære), 2751 arna (MS arra), Exodus 321 leon (MS leor), Elene 1183 foran (MS fonan), Christ 921 gehwone (MS gehwore), Phoenix 336 gehwone (MS gehwore), and others.

63 Sedgefield suggested emending the problematic ealuscerpen in line 769 to ealscerpen; which, if accepted, would be further evidence for the scribal confusion in question. But no later editor has adopted Sedgefield's emendation, and in any case I have attempted to base my argument on instances where there is general agreement among editors about the need for correction, rather than on random conjectures by individual editors.

64 Cf. Campbell, , Old English Grammar, § 60.Google Scholar Our letter w is an importation by Norman scribes, and is not found in pre-Conquest manuscripts.

65 On the letter wyn(n) in futhorc, see Page, R. I., An Introduction to English Runes (London, 1973), p. 78Google Scholar, and, more recently, Parsons, D. N., Recasting the Runes: the Reform of the Anglo-Saxon Futhorc (Uppsala, 1999), pp. 26–7 et passim.Google Scholar

66 The Epinal, Erfurt, Werden and Corpus Glossaries, ed. Bischoff et al.: 96r pestsuppind [A 527], 96v eastnorppind [B 119], 97r pindil [C 26], 107r pand [T 97], 170v þinþindae [U 19], 107v edpalla [U 31] and 107v paeffsas [U 35].

67 As for example by the scribe of the ‘Corpus Glossary’, writing at Canterbury c. 800: see The Epinal, Erfurt, Werden and Corpus Glossaries, ed. Bischoff et al., 4r permod (A 8), 4v suðanpestan (A 88), praec (A 132), 5r gesiupede (A 177), 5v praene (A 227), to peorðmyndum (A 229), etc.

68 The continental scribe of the ‘Erfurt Glossary’, copying from an exemplar very similar to, but not identical with, the ‘Epinal Glossary’, rendered the gloss pindil as pindil: see The Epinal, Erfurt, Werden and Corpus Glossaries, ed. Bischoff, et al. , and Pheifer, J. D., Old English Glosses in the Epinal-Erfurt Glossary (Oxford, 1974), p. 11Google Scholar (no. 173). Similar confusions pervade Old English texts: for example, the scribe of the Vercelli manuscript rendered ofer sponrade of Elene 996 meaninglessly as ofer span rode; cf., in the Exeter Book, Gutblac 875 stopum (MS stopum), Juliana 294 bispeop (MS bispeop) and, in the Junius manuscript, Christ and Satan 318 hreopan (MS hreopan).

69 Cf. Charters of Selsly, ed. Kelly, S. E., Anglo-Saxon Charters 6 (London, 1998), 74.Google Scholar

70 In references to charters here and throughout, S = Sawyer, P. H., Anglo-Saxon Charters: an Annotated List and Bibliography (London, 1968).Google Scholar The present charter (S 65) is preserved as BL, Cotton Augustus ii. 82, and ptd de Gray Birch, W., Cartularium Saxonicum, 3 vols. and index (London, 18851899Google Scholar) [hereafter BCS], no. 11; it is ed. in facsimile in Bond, E. A., Facsimiles of Ancient Charters in the British Museum, 4 vols. (London, 18731878Google Scholar) [hereafter BM Facs.] I, no. 3, and illustrated in Bruckner, A. and Marichal, R., Chartae Latinae Antiquiores, III: British Museum, London (Olten, 1963)Google Scholar [hereafter ChLA[, no. 188.

71 S 90: Canterbury, Dean and Chapter, Ant. M. 363; ptd BCS, no. 162, and ed. in facsimile in Sanders, W. B., Facsimiles of Anglo-Saxon Manuscripts, 3 vols. (Southampton, 18781884) [hereafter OS Facs.] I, no. 1.Google Scholar

72 Listed Gneuss, Handlist, no. 25, and CLA II, no. 139; ed. in facsimile by Blair, P. Hunter and Mynors, R. A. B., The Moore Bede, EEMF 9 (Copenhagen, 1959).Google Scholar

73 Lasted Gneuss, Handlist, no. 173, and CLA II, no. 133; also illustrated in Keynes, , Anglo-Saxon Manuscripts in Trinity College, pp. 89 and pl. I.Google Scholar

74 Further examples may be seen in CLA II, nos. 184,189, etc.

75 To these instances should probably be added the four instances where Scribe I spells the word maðum with d rather than with ð: 385 (madmas), 472 (madmas), 1198 (bordmadmum), 1528 (madme); and in the stint of Scribe II note in line 2869 MS pryðlicost, which Thorkelin and others have corrected to pryðlicost. Cf. the comments of Hoops, J., Beowulfstudien (Heidelberg, 1932), pp. 7 and 136.Google Scholar

76 Listed Gneuss, Handlist, no. 626; ed. in facsimile by Ker, N. R., The Pastoral Care: King Alfred's Translation of St Gregory's Regula Pastoralis, EEMF 6 (Copenhagen, 1956)Google Scholar, and also illustrated in Keller, W., Angelsächsische Palaeographie, 2 vols. (Berlin, 1906), pl. III.Google Scholar

77 In such contexts they also used the originally runic letter thorn (þ); see above, p. 24.

78 On one occasion the scribe of the ‘Epinal Glossary’ did so himself, as may be seen in die gloss pictus acu mið næðlae asiuuid (‘sewn with a needle’) [103r: P 247]. This is apparendy the earliest example of the letter ð in Anglo-Saxon script.

79 S 19: BL, Stowe Ch. 1, ptd BCS no. 97, and ed. in facsimile in OS Focs. III, no. 1, and ChLA, no. 220.

80 S 21: BL, Cotton Augustus ii. 88: ptd BCS 98, and ed. in facsimile in BM Facs. I, no. 4 and ChLA, no. 189.

81 S 8: ptd BCS no. 45, and ed. in facsimile in BM Facs. I, no. 1 and ChLA, no. 182, as well as Lowe, E. A., English Uncial (Oxford, 1960), pl. XXI.Google Scholar

82 S 24: BL, Cotton Augustus ii. 101, ptd BCS no. 160, and ed. in facsimile in BM Facs. I, no. 8 and ChLA, no. 192.

83 S 106: BL, Cotton Augustus ii. 26, 27: ptd BCS no. 201, and ed. in facsimile in BM Facs. I, no. 9 and ChLA, no. 186.

84 S 153: BL, Cotton Augustus ii. 97: ptd BCS no. 289, and ed. in facsimile in BM Facs. I, no. 12 and ChLA, no. 191.

85 E.g. the ‘Oslac’ charter, securely dated to 780: S 1184, ptd BCS no. 237, and ed. in facsimile by Keynes, S., Facsimiles of Anglo-Saxon Charters, Anglo-Saxon Charters Supplementary Volume 1 (London, 1991), no. 2 and pl. IIGoogle Scholar (the names Alðheard, Æðelmund, Beorhtnoð, etc.).

86 Peter Clemoes pointed to a single example of open a in a manuscript from the beginning of the tenth century (Interactions, p. 33, n. 77), illustrated in Dumville, ‘English Square Minuscule Script: the Background and Earliest Phases’, pl. I, with discussion ibid. p. 166, n. 96. See also Emms, R., ‘The Scribe of the Paris Psalter’, ASE 28 (1999), 179–83, at 180Google Scholar, where he points to isolated examples of what he calls ‘open topped a’, illustrated in his pis. VIa and VIb. In these cases, one swallow does not make a summer.

87 On the Thorkelin transcribers, see now Gerritsen, J., ‘What Use are the Thorkelin Transcripts of Beowulf?’, ASE 28 (1999), 2342.Google Scholar Cf. also lines 1735 and 2008.

88 Andrew, , Postscript on Beowulf, p. 133Google Scholar: ‘there are over sixty words in which a letter is dropped or inserted or changed; while many of these are obvious slips of the pen, a fair number are misspellings which might destroy or distort the sense, e.g. 447 deore = dreore, 84 secg = ecg, 1529 hord = hond, 2771 wræce = wræte’. If one eliminates the cases where a letter has simply been omitted - assuming that the omission is not a result of literal confusion but merely of carelessness — then the total number of errors in the transmitted text which result from literal confusion is smaller still, and the percentage of such errors which can be explained by my hypothesis of an archetype in set minuscule, correspondingly greater.

89 There is little point in listing here those scholars who have previously accepted an early-eighth-century date for Beowulf, and those who have not; see, in general, the balanced assessment of Liuzza, R. M., ‘On the Dating of Beowulf’, in Beowulf: Basic Readings, ed. Baker, P. S. (New York, 1995), pp. 281302.Google Scholar

90 Lapidge, M., ‘Beowulf, Aldhelm, the Libermonstrorum and Wessex’, SM, 3rd ser. 23 (1982), 151–92, esp. 188–9Google Scholar (repr. in Angio-Latin Literature 600–899 (London, 1996), pp. 271–312, at 308–9).

91 Fulk, R. D.,. A History of Old English Meter (Philadelphia, PA, 1992), p. 390Google Scholar: ‘In brief Beowulf almost certainly was not composed after ca. 725 if Mercian in origin, or after ca. 825 if Northumbrian. These findings confirm and support the relative chronological evidence that it is one of the earliest poems in English. Whether Beowulf is Northumbrian or Mercian in origin cannot be determined with assurance, but what evidence there is suggests that it is Mercian.’

92 Cf. Clemoes, , Interactions, p. 34Google Scholar: ‘there is a prima facie case for supposing that our poem underwent a three-stage textual transmission, consisting of a pre-mid-ninth-century writing, a post-mid-ninth-century copying, and lastly another, very early-eleventh-century, one.’

93 The features in question are set out concisely by Klaeber, , Beowulf and the Fight at Finnsburg, pp. lxxilxxxviii, passim.Google Scholar

94 O'Keeffe, K. O'Brien, Visible Song: Transitional Literacy in Old English Verse, CSASE 4 (Cambridge, 1990).Google Scholar On the implications of her arguments for textual criticism, see Fulk, R. D., ‘Inductive Methods in the Textual Criticism of Old English Verse’, Medievalia et Humanistica 23 (1996), 124, at 7Google Scholar: ‘if this argument [scil. O'Brien O'Keeffe's] is correct, it is probably useless to set about restoring an original text, since the degree of rewriting involved in the scribes' work suggests that the surviving poetry represents layered compositions in which the seams cannot be detected with assurance’.

95On the Dating of Beowulf’, pp. 292–3.

96 As when he stated at an earlier point that Hygelac had given Beowulf ‘seven thousand’:

þæt he on Biowulfes bearm alegde

ond him gesealde seofan þusendo.

bold ond bregostol. (2194–6)

97 I am very grateful to Andy Orchard for help with the homileric language in Hrothgar's ‘sermon’, and for drawing my attention to relevant parallels.

98 The Homilies of Wulfstan, ed. Bethurum, D. (Oxford, 1957), p. 258.Google Scholar

99 The Vercelli Homilies, ed. Scragg, D. G., EETS os 300 (Oxford, 1992), 60.Google Scholar

100 The Blickling Homilies, ed. Morris, R., EETS os 58,63 and 73 (Oxford, 18741880; repr. as one vol., 1967), 95.Google Scholar

101 Eleven Old English Rogationtide Homilies, ed. Bazire, J. and Cross, J. E. (Toronto, 1982), p. 113.Google Scholar

102 In light of the parallels from sermons, especially that in Rogationtide VIII (on ece oððe on adl), Matti Kilpiö suggested to me that ecg (“sword”) in Beowulf 1763 (þæt þec adl oððe ecg eafoþes getwæfeð) should probably be emended to ece (‘ache’, “pain”), inasmuch as the affliction caused by the sword is mentioned in Beowulf a few lines later (1765: oððe gripe meces). If such a conjecture were adopted, it would have the implication that it was not Scribe I who made the interpolation, but that he misunderstood the form ece in an earlier exemplar.

103 Whether the poem ever existed in an oral form cannot be known; the most that can be said is that stories about Beowulf and a monster named Grendel may have been in oral circulation before the poem was composed in writing (cf. Lapidge, , ‘Beowulf, Aldhelm, the Liber monstrorum and Wessex’, pp. 179–84Google Scholar (repr. Anglo-Latin Literature 600–899, pp. 299–304)). I therefore dissent from the views of earlier scholars such as Prokosch, E., ‘Two Types of Scribal Errors in the Beowulf MS’, in Studies in English Philology: a Miscellany in Honor of Frederick Klaeber, ed. Malone, K. and Ruud, M. B. (Minneapolis, MN, 1929), pp. 196207Google Scholar, who suggested that Scribe I was writing from dictation, whereas Scribe II was simply copying from a written exemplar; and Sisam, , The Structure of Beowulf, p. 67Google Scholar, who as one of two alternatives suggested ‘that the poem was composed without writing, and recited from memory by trained entertainers until it was recorded, perhaps at the request of a king or noble who heard and admired it, and wanted to be sure that he could hear it again in private or in company’. I favour the other alternative mooted by Sisam (ibid), ‘that the poem was written down by the poet… before it was made known to readers or listeners, and thereafter was transmitted in writing’.

104 The argument was first made by Sisam, , Studies, pp. 93–6Google Scholar, and has been set out in greater detail by Orchard, Pride and Prodigies: see pp. 26–7 et passim.

105 Three Old English Prose Texts, ed. Rypins, , pp. 6876.Google Scholar Rypins's apparatus is poorly and confusingly laid out, and it is very difficult to distinguish at any point between manuscript readings and readings printed by earlier editors. (The text cries out for a new edition.)

106 Based on the apparatus criticus of Orchard, Pride and Prodigies, pp. 184–200. In one case which may affect the count, the reported reading underbreðað has not been assigned to a manuscript source (ch. 21, n. K), but is in fact in V.

107 Cf. Sisam, , Studies, pp. 93–4Google Scholar: ‘We began with the supposition that the first and last pieces, Christopher and Judith (which themselves show no signs of having had a manuscript transmission in common), differ in transmission from the remaining three, Wonders, the Letter, and Beowulf. This fits all the evidence. But although these last three texts are fairly old, no evidence has appeared to support the hypothesis that they, or any two of them, make an old nucleus for the collection.’

108 To cite one example: in the ninth-century interlinear gloss to the Vespasian Psalter, there are large numbers of cases where d is written in places where we would expect ð (in words such as eord for eorð, or dætte for ðætte, etc.). In light of the recent demonstration that the Vespasian Psalter gloss must be a word-for-word copy of an earlier interlinear gloss (see Pulsiano, P., ‘The Originality of the Old English Gloss of the Vespasian Psalter and its Relation to the Gloss of the Junius Psalter’, ASE 25 (1996), 3762Google Scholar), one wonders whether the many confusions of d and ð might imply that this hypothetical earlier gloss should be dated before c. 750. But such a question could only be answered by a thorough study of the transmitted form of the gloss. Note also that there are numerous examples of confusion of d and ð throughout the corpus of Old English poetry, as even a cursory reading of the apparatus criticus in ASPR makes clear.

109 A preliminary version of this article was read to a meeting of the International Society of Anglo-Saxonists at the University of Notre Dame in August 1999, and I am very grateful to members of the audience, particularly Carl Berkhout, Mechthild Gretsch, Matti Kilpio, Katherine O'Brien O'Keeffe and Andy Orchard, for suggesting ways in which the argument required revision. In preparing the article for publication, I have had the benefit of the philological expertise of Mechthild Gretsch, who saved me from many errors. As always, Simon Keynes provided sure guidance through the charter evidence. I owe a long-standing debt to Andy Orchard, with whom I shared the pleasure of teaching Beowulf in our Cambridge seminars for a number of years, and who has always generously shared his knowledge of the poem with me.