University of Virginia Library

The Rôle of Formulas in the Dissemination of a Middle English Alliterative Romance
by
Hoyt N. Duggan

We do not know very much about the dissemination of vernacular romances in fourteenth and fifteenth-century England. What we do know is fragmentary, derived in bits and pieces from accounts, letters, wills, and random notes in manuscripts. There is some evidence for commercial scriptoria in mid-fourteenth-century London in which English romances were copied and perhaps composed and more evidence after the turn of the century.[1] We know too from various sources that at least some aristocratic and middle class families either retained or hired on a casual basis scriveners who copied literary texts.[2] Nevertheless, our knowledge in this important area is severely circumscribed, especially in the interreaction of oral performance and manuscript copying as dual means by which literary

texts were presented to the public. Obviously the stationers, the monastic and secular scriptoria, the amateur scribes, and independent scriveners like William Ebesham or Chaucer's "owne scriveyn," Adam, must account for the vast bulk of surviving literary manuscripts. For manuscripts descended in this kind of tradition the methods of textual criticism as developed in the treatment of biblical and classical texts are adequate and relatively straightforward. However, in a culture in which literature was written to be read or recited aloud and in which a special class of entertainers existed to satisfy a demand for orally presented narratives, it would be surprising if some of the extant texts were not touched in important ways by the facts of oral recitation, especially in the intrusion or rearrangement of materials by the performers.

Professor Albert C. Baugh has recently demonstrated just such a process in the surviving manuscripts of three Middle English romances, Beves of Hampton (7 manuscripts), Guy of Warwick (6 manuscripts), and Richard the Lion-Hearted (8 manuscripts). As Baugh noted, "the variations between the manuscripts containing these longer narratives are often quite striking, so great that it is generally impossible to establish a single critical text with the variants adequately reported in the footnotes."[3] Late nineteenth-century editors had attempted to account for these variants in terms of scribal sophistication, of conscious attempts to improve the text. But as Baugh notes, these apparent instances of sophistication are not just a few touches here and there but extensive throughout the romances. Moreover, they represent "improvements" of a very curious nature, since most frequently, the improvement is inferior to the reading attested by the other manuscripts. Changes of this sort tend, in Baugh's words, to be "banalities, clichés, and rime-tags." The simplest and most compelling explanation for these variants lies in the combination of transmission by manuscript copying and the improvisation of oral performers. "In short," Professor Baugh concludes, ". . . for those popular romances in which the manuscripts show extensive variations of a kind that cannot be explained as scribal errors we have in each case the story as told by some minstrel."[4]

The implications for editors of Middle English romances of this kind surviving in more than one manuscript are perfectly clear, though Professor Baugh does not insist upon them: editors must in these cases present parallel-text editions or choose a best text to be printed save where its readings produce obvious nonsense or may be attributed to scribal error.[5] Elaborate comparison of the variants is, to a considerable extent, an exercise in futility, since each manuscript is not simply a more (or less) authoritative

voice representing the poet's original text as changed and corrupted in a process of mechanical copying but each may well be the result of a performer's collaboration with the poet. Each manuscript is an authority only for a performance or series of performances lying behind it.

However, the introduction of performers' variants does not, in every case, wreak such havoc with the manuscript tradition, and there is some evidence, in alliterative poetry at least, that a happy conjunction of a performer's aural memory and a good exemplar provided a remarkably fine basis for copying a longish romance. The romance in question is The Wars of Alexander, an alliterative poem translated from the I3 recension of the Historia de preliis Alexandri Magni sometime after 1360 in the Northwest Midlands. The Wars survives in two fragments. The older and longer of the two, MS. Ashmole 44, of the Bodleian Library (MS. A henceforth), was written in a northern dialect about the middle of the fifteenth century. Its ninety-seven folios, comprising 5677 lines, preserve all of the romance except the closing episode describing the death of Alexander. The shorter manuscript, MS. Dublin D.4.12, Trinity College (MS. D), was written in or near Durham Priory in the last quarter of the fifteenth century. It lacks the first 677 lines and ends at line 3425, but for 2686 lines, it provides a check against the A manuscript. W. W. Skeat presented a parallel-text edition of the two manuscripts in 1886 in which he occasionally offered emendations but did not offer an opinion as to the possibility or desirability of establishing a critical text by analysis of the variant readings. His evidence for emendation was for the greatest part based on paleography, metrical irregularity, or reference to his unfortunately corrupt edition of the Historia de preliis.[6]

When I decided to re-edit the poem, initial comparison of the two manuscripts led me think it possible to establish a critical text by comparison of the two manuscripts and by reference to the source, that is, by the traditional methods of editors. However, on the way to establishing a critical text, several interesting complications presented themselves, problems of more interest than the text of The Wars of Alexander itself. Solution of those problems throws some light upon the question of how oral performances affected the dissemination of literary works and provides a model for the use of formulaic diction in deciding between variant readings in the manuscripts.

It is not immediately obvious that either manuscript has been touched in any way by oral improvisation. Skeat explicitly denied that the A manuscript showed any sign of oral transmission (p. xiv), and as I shall suggest

below, there is the firmest kind of evidence to link the two manuscripts to the original by continuous lines of physical copying. Nevertheless, analysis of the variants between the two manuscripts and of metrical irregularities when only the Ashmole text presented readings indicates a strong probability that both manuscripts have readings which can best be explained by recourse to minstrel improvisation.

Many variations between the two texts are the result of mechanical errors in copying of one kind and another—repetitions, transpositions of letters, confusion of graphs, omissions of words or lines, and the other unconscious errors that plague anyone who attempts to copy a lengthy text. There are also apparent instances of scribal sophistication, of deliberate attempts of both scribes to edit the text and improve it. But after one has added all instances of deliberate scribal sophistication to the variants arising from mechanical miscopying of the manuscript, the bulk of the variants is still to be explained. There are passages in which one text is in the present tense, the other in the preterite (e.g., at ll. 885 ff., 921-924, 943-946, 1024 ff., 1142-1146, 1385-89, 1397-1406, etc.), in which active and passive voice vary, or in which mood and number differ (e.g., ll. 722, 844, 1944, 2142, 2164, 2987, 3095, etc.). In most cases, either reading is suitable to the context. Even more striking are variations in lexical items: stith] styffe, ll. 1069, 1251, 1327, etc.; taite] ioy, l. 1208; ryfe] gret, l. 1352; bild] tild, l. 1366; tellis] wittnes, ll. 1439; 1592; dere] derfe, ll. 1024, 3013; droune] drench(yd), ll. 2590, 3072, 3274; wathe] wound, l. 1411; fone] some, l. 3180; man] tulke, l. 3174; douth] doughty, ll. 2627, 2663, etc.; worthili] wightly, ll. 1405, 1428; floum] flode, l. 2898; enmy] foes, l. 3096. The list of such variants can be easily expanded by anyone who inspects even casually the manuscripts. Some of them can be explained as dialect translation in which the scribe substituted a more common word for a distinctively Northern form, as in the taite] ioy; fone] some variation. Others, like man] tulke; floum] flode; tellis] wittnes; enmy] foes, etc., are probably to be explained as the more or less unconscious substitution of synonyms. But there exists a kind of systematic variation between the two manuscripts that appears to be beyond the normal and usual combination of mechanical errors, unconscious substitutions, and scribal sophistication. This demands some explanation.

The language of the poem is highly formulaic; that is, there is considerable repetition of half-lines within the poem, with both identical collocations and formula systems in evidence. There is every reason to think that the poet was literate, that he composed his poem at his table with a Latin manuscript before him, and no reason to consider him an oral poet. Yet the poem is at least as formulaic as Beowulf or Morte Arthure and shares most characteristics of orally composed poetry. Those characteristics are, I believe, fairly generally understood, having been much discussed in recent years, and it is not necessary to recapitulate the theory in detail here.[7]

Suffice it to say that the poet worked within a relatively limited subgrammar of the language with conventional phrases which lent themselves to preserving a limited set of metrical and rhythmic patterns but which could be almost endlessly extended through use of what R. A. Waldron called "rhythmical-syntactical moulds."[8] The existence of these formulas and formula systems is, as I shall argue below, one of the primary critical tools for establishing the probable originality of many manuscript variants, since the variant that corresponds to a well-attested formula or formula system is more probably original than a reading that fails to fit within the system. However, there are several sets of formulas which exhibit a surprising (if one thinks only of error or scribal sophistication) consistency in variation between the two manuscripts, even when the formulas concerned are separated by several folios. The most striking of these are listed below with the A reading in italics and D inserted in parentheses:

• 1. And passis (passyd) on to Persy . . . 2029
• He passis (passed) on toward Persy . . . 2130

The key element in this case may be the word pass in the first stressed and alliterating position, since the variation also occurs in different formula systems:

• þan passe (passyd) vp oure princes . . . 1397
• þan passes (passid) he þethen with his princes . . . 1076
• Passis (Passyd) into þe palais . . . 3217

---

270




• 2. . . . out of þe west endis (ende) 1733
• . . . into þe west endes (ende) 2325
• 3. . . . was Wyothy hatten (haldyn) 2150
• . . . Mocian was hatten (haldyn) 2540
• . . . my satroparis hatten (halden) 1913

For other regular instances of this formula system, see ll. 40, 836*, 914, 1913, 2037, 2106, 2297, 4720, 5093, etc.

• 4. Aires (Kayres) him to ser Alexander . . . 2637
• Aires (Kayres hym) to ser Alexander . . . 2680
• Aires (Karyn) þaim to ser Alexander . . . 2792
• þen aires (kayrez) him on ser Alexander . . . 2846
• þare aires (kayres) him in ser Alexander . . . 2918
• Aires (Caryn þaim) to ser Alexander . . . 3110
• And aires (karys) with þaim to Eufraten . . . 2770
• þan aires (kayres) he with Emynelows . . . 3005
• 5. (I,) Alexander þe aire (D omits þe) . . . 1838, 2319
• 6. þan Alexander (als) belyue . . . 1096, 1425, 2209, 2886, 3117
• And Alexander (als) belyfe . . . 909, 956, 1792, 2936
• To (ser) Alexander (als) belyue . . . 1255, 3084, 3181
• Quod Alexander (als) belyue . . . 1684, 2271

In this instance, D is almost certainly correct, since A agrees with D in reading as at ll. 710, 1957, 2183, 3029, and 5197.

• 7. Aires (hym) into Affrike . . . 1050
• þan aires (hym) on ser Alexander . . . 2265
• Aires (hym) to ser Alexander . . . 2680
• Aires (þaim) to ser Alexander . . . 3110

Again, D is probably correct, since A agrees in reading him at ll. 749, 2847, 2918, etc.

• 8. . . . he flittis with his ost (D omits with) 2173, 2439

In this case, A is probably correct. Cf. ll. 2266, 3184, 3207, 3949, 4067, and 4787. A possible counter instance in A is l. 3784.

9. Authority tags occur some eighty-five times in the b-line with variant forms of the pattern "as þe buke tellis," "as me þe claus tellis," "as demys þe textis," "þe scriptore it callis," "þe romance it tellis," "as I am enfourmed," etc. These tags alliterate variously on vowels and the following consonants: /b,d,f,k,j,l,m,p,r,s,t,w/. There are thirty-one lines in which both manuscripts agree and are regular,[9] thirty instances in which A is the

sole authority and alliteration is regular,[10] one instance in which A is regular and D is not (l. 3389), ten instances in which A fails to alliterate and is corrected by D, and another eight instances in which A is the only manuscript and fails to alliterate regularly. Only two instances occur in which both manuscripts agree in irregular alliteration (ll. 2233, 3372). Of the forty-five instances of this formula occurring where both manuscripts offer readings, forty-three (95.6%) are regular. Two instances in which A is irregular and corrected by D may be explained as the scribal substitution of the synonym tellis for the correct neuens and says me of D (ll. 1485, 2997). Of the remaining eight instances in which D may be compared to A, D is regular and A in every case reads "as þe buke sais~tellis~witnes."                  

Dublin MS.	Ashmole MS.
als says me þe writtes	as þe buke sais	881
als says me þe text	as þe buke witnes	916
as menys me þe writtes	as þe buke tellis	1615
as I am enformed	as þe buke tellis	1691
as sayn me þe writtes	as þe buke sais	2112
& þus says þe text	as þe buke sais	2536
as þe writt shewys	as þe buke tellis	2539
as demys me [þe] writtes	as þe buke tellis	3069

In the remaining passages where the D text is lost, every instance in which A fails to alliterate regularly reads "as þe buke tellis~sais" (e.g., ll. 17, 203, 3789, 3888, 3890, 4201, 4721, and 5479.

In addition to this clearly formulaic variation, there are regular word substitutions which, though not part of identical formulas, occur in the same or similar metrical environs. There is, for example, consistent variation in line terminal position of horses] blonkez, ll. 791, 2156, 2399, 2981; sendis] wayfez, sende] wayfe, ll. 1716, 1868 (cf. 2752); callis] clepys, ll. 2184, 3117; ost] (h)ostez, ll. 1115, 2031, 2173, 2266, 3207 (and within the line at 3111). There is consistent variation at other positions within the line: out] forth (after verbs of motion), ll. 802, 931, 948, 1170, 1218, 1243, 1318, 1394, 1653 (but cf. l. 923); him] omitted by D, ll. 1078, 1113, 1196, 1353, 1451, 1955, 2283, 2594, 2610, 2662, 2956, 2988, 3114, 3190; meth] might, ll. 816, 1981, 3102; gome] grome, ll. 1190, 1936, 2101; douth] doghty, ll. 2627, 2663, 3006, 3061, 3157; tilt] typed, tiltis] typys, ll. 1303, 1418; worthili] wightly, ll. 1405, 1428; quen] whil(s), ll. 2172, 2255, 2806; stith] styffe, ll. 1069, 1251, 1327, 2050; stithli] styfly, l. 2429; auncient] auancet, ll. 1002, 2391; and faithely] faythfully, ll. 3175, 3282. This remarkable consistency of variation suggests a systematic approach to revision rather like that of a modern editor, or else the habitual formulaic or semi-formulaic collocations of someone who knows the text very well by memory, possibly a performer.

R. F. Lawrence has recently pointed to another remarkable aspect of the manuscript variants in the poem, and that is the tendency for the differing readings of both manuscripts to be metrically regular, each reading conforming to one of a limited number of rhythmic norms.[11] I list below a few instances of this sort in which the same rhythmic pattern occurs in both manuscripts in spite of differences in wording, in tense, or in number. Unstressed syllables before the first stressed stave are assumed not to be significant in determining the rhythmic type, and final -e is sounded only at the end of the line, while -id and -is are syllabic when they follow a stressed syllable.[12] The differing syllables are in italics.

                                       

Ashmole MS.	Dublin MS.
in his brath endis	whilse hys breth lastez	1220
many lede floȝen	in mony lowd showte	1392
encumbrid þaim neuir	he comber þaim neuir	1480
he quirys all-to-gedire	enquirez all-to-gedyr	1703
Sir Darius him-seluyn	indited you hym-seluen	1823
& gefe vs ȝour lefe	& gyfe vs owr lyfez	1826
þofe he wele suffir	of þe whele sofre	1858
na vaunte sall arise	no vaunt sall þar rise	1880
ȝow limpid to encumbre	enlympyd you to combre	1881
& besely we shapid	& besely echapyn	1944
folke to be nombrid	& folkez unnowmyrd	1992
þe princes to schewe	þe prince it to shewe	2029
& swyth þus him tellis	& sothly hym tald	2069
we wast þam for euire	we wastyd for euer	2330
with a kene voice at anys	& cried all at onys	2345
& ernstly he spekis	& egirly spekes	2348
fulfillis his will	fulfyll þan hys wylle	2359
ȝoure conscience it opence	in your conscience doys shew	2422
his cors for to bathe	hys Cors to be bathyd	2542

The list might be extended beyond this 1300 line sample, and within the sample the list would be perhaps tripled if one were to include variants which were metrically regular though not necessarily identical in form or in which only one word changed.                

to kepe at þam fall	þam paim to kepe fallez	1192
& brathly woundid	& wykydly þaim woundes	1214
fey to þe gronde	fast to þe grunde	1215
brymly he smytis	brathly he smytez	1222
glidis fra othire	glydes fast þair way	1310
feȝtand ȝerne	feghtyng full ȝarne	1315
to of his turnes rekyn	his tournays to reken	1404
of þe shire son	shott fro þe son	1544

Lawrence credits such metrical regularity to the scribes: It is not impossible that the variants of this sort are purely scribal, for all of the evidence suggests that many scribes made free with their exemplars, changing texts more or less freely to make them coincide with their aesthetic and (sometimes) moral or political preferences.[13] However, the best evidence available in George Kane's study of scribal tendencies in the sixteen manuscripts of the A version of Piers Plowman does not appear to support the idea of consistent and systematic variation of the sort we have discovered here. Kane wrote: The kinds of consistent variation in lexicon, in formulas, and in rhythmic structure in The Wars of Alexander are perhaps best explained by a limited form of oral transmission.

That the poem was intended for oral recitation the text itself shows clearly. More importantly, both of the surviving manuscripts show in their indications of divisions the impact of oral performances. As Albert C. Baugh pointed out in an important study of the romances in Middle English, The Wars of Alexander is divided into units of from 120 to 312 lines, thus allowing natural breaks for performers.[14] The poet was careful to provide indications of convenient places to break the narration at points greater than the single passus, usually with a formulaic line or two:

The ends of Passus IV and V are not so marked, but VI and VII are (ll. 1578-1579, 1842-1843), and Passus VIII ends with no distinctive formula but introduces the letter which will begin the next passus. The pattern is resumed at the end of Passus IX: Significantly, these lines are not in D, nor are they part of the normal twenty-four line strophe.[15] The second line is unmetrical as it stands, whereas the first repeats almost verbatim the preceding formulaic conclusion at l. 1718. All of the evidence suggests that the two lines are not authorial, for the poet in all other instances worked the transitional formula into the normal strophe. The poet had not provided such a formal break before the end of Passus XI at l. 2845, leaving an uninterrupted text of about four times the length of those above it, and it would appear that the lines were attached, inelegantly, by a performer who had found it necessary to stop his recitation or reading at that point.[16] The poet does not provide another such ending at the close of Passus X or XII, but does close Passus XI and XIII with the laconic half-line "and here a passe (fitt) endes." Interestingly, D takes no notice of A's Passus XI and XIII, so that it has twelve passus to A's fourteen. Probably the passus divisions in A are authorial, since D's Passus X of 528 lines is half again as long as the next longest passus. One explanation, admittedly no more than conjecture, for the loss of passus headings in MS. D is that the passus breaks at these points were not useful to the performer-scribe, who subsequently dropped them from his text.[17]

There is every reason to believe that the formulas cited above were intended to meet the exigencies of oral recitation—the wandering attention of an audience, the other scheduled entertainment, or the mere improbability of reading through the entire poem in one or two sittings.[18] Phrases

like "And if ȝow likis of þis lare to lesten any forthire" are not meaningless rhetorical questions but are intended to allow a performer to determine the desires and interest of his audience and thus to decide whether he should continue or not. As such, they are the clearest kind of evidence that the poet wrote with an eye to the essentially oral nature of any given performance.

Larry D. Benson is almost certainly mistaken in asserting that "clearly such works as The Wars of Alexander and The Destruction of Troy were written primarily for readers."[19] Few instances of alliteration are intended to appeal to the eye, but there are a significant number of instances in which elision alliteration occurs, in which only the oral performance makes the alliterative pattern clear, as for example in l. 1829, where elision of at and ese produces the /t/ necessary for regular alliteration: "Takis þam with him to his tent & þam at ese makis."[20] Moreover, the poet consistently appeals to an audience of auditors. He closes the first half of the poem with an appeal which makes sense only in terms of a listening audience and opens the second half of the poem with what amounts to a formal reintroduction:

There is a curious ambivalence in these lines toward the poem and the creative process which produced it, for one of the poet's concerns is that the performer appear to be thinking and composing on his feet, that he is not sitting at his desk with pen in hand but composing ex tempore and finding it necessary to "tary for a time & tempire my wittis." The other interest, submerged to this point, is that his audience should recognize the story as history, that he, the poet, has indeed followed the letter and that the poem is already "made." The prayer for God's blessing is for the maker, who need not be the person who recites the romance, and in this one passage the poet, speaking as suppliant and translator, steps for a moment outside the narrative persona which in all other instances is consonant with the performer-composer. Here we see most clearly the double perspective maintained by a literate poet writing within an essentially oral tradition for oral delivery. Perhaps the same tensions between oral and written language as modes of dissemination can best account for the kinds of variation we have encountered in the two surviving manuscripts.

The researches of Albert C. Baugh into the authorship and dissemination of Middle English romances support such a conclusion, providing considerable evidence of romances changing under the influence of dissemination through minstrels.[21] It would have been no unusual feat for a minstrel to memorize the 6000-plus lines of The Wars of Alexander. Léon Gautier cites instances of French jongleurs boasting of having entire chansons de geste by memory and being able to recite them without losing a line.[22] But the performer who recites a poem of this length a number of times must sooner or later begin to change it in larger and smaller details, substituting his own habitual collocations for those of the text he had once conned so carefully. It is the practicing performer rather than the scribe who is likely to produce the kind of variants we have noted above, for it is he who by countless recitations of both this and other alliterative poems will have fixed in his mind not only the formulas but also the abstract grammatical and metrical patterns that lie behind them. The very slowness of the pace in ordinary scribal transcription, the necessity for copying word by word, or at most phrase by phrase, does not encourage success at imitating alliterative formula systems.[23]

There are, nevertheless, serious difficulties in the way of one who would argue that either the A or D manuscript is separated from the original by oral recomposition. The experience of Milman J. Parry and Albert B. Lord with modern Slavic singers of tales contradicts the textual situation in this instance. Although the modern oral singers boasted of their ability to repeat a lengthy tale from memory without losing a line, they did not in practice often give a word for word, or even a line for line identical version of the tale.[24] There is, however, no compelling reason to assume

that the conventions governing a primarily oral art like that of the modern Slavic singers would be identical with those for the propagation orally of written texts. The memory of a minstrel reciting a Middle English romance is probably better compared to that of the modern actor, without, of course, the actor's respect for "established" text, than with the memory of the singer who composes orally. The formula may serve as an aid to memory as well as provide the basic patterns of a grammar of composition, and it is to be expected that in a transitional text like The Wars of Alexander formulas would function to aid in the dissemination of the poem as much as it would have aided the composition.[25]

A more forcible objection to a theory of oral transmission is that both manuscripts are demonstrably linked by written tradition to the original; that is, marks in the manuscripts correlate both the strophe divisions and major rubrications in the Latin source of the poem, neither of which would be possible in a purely oral tradition. In the A text, there are fifty-nine instances in which the scribe marks the beginning of a strophe with '¶', while the D text has fifty significant instances of lines either marked with '¶' or written wholly or partially in majuscule letters. Of these, there are

seventeen instances of agreement between the two manuscripts which occur at the beginning of a strophe and one instance (l. 2755) in which both manuscripts agree with a major Latin manuscript division which does not correlate with the usual twenty-four line strophe. Only a continuous line of copied texts will account for the presence of these division markers.[26]

We may consider the question of why a performer would go to the trouble and expense of copying a text he had memorized. Unfortunately, too little is known of the actual process by which performers acquired copies of literary works. As Dieter Mehl pointed out, there is little evidence "in favour of minstrel-collections, i.e., manuscripts written specially for minstrels and carried around by them."[27] On the other hand, it is clear that there was a class of performers, some itinerant, some more comfortably attached to a court and patron, who had to have such literary materials for their livelihood. That itinerant minstrels could have afforded any sizeable collection of manuscripts is improbable, but their more favorably established confreres might well have been encouraged to obtain and copy literary works. Performers willing to write out and sell copies of manuscripts in their own, or more probably, in their patron's collection, cannot have been rare. Clearly there was a market for such manuscripts, for in a much quoted passage, Petrarch decries the existence of a sub-artistic class of minstrels, a species of literary parasite who built their reputations upon the purchased or begged wits of original poets:

As Baugh remarks, it is probable that the same situation obtained in fourteenth-century England.[28]

Oral transmission of the sort Professor Baugh found is out of the question for this poem, and any simple notion that the texts are the product of oral improvisation requires strict qualification. Still, there is the remarkable consistency of the manuscript variants which must be accounted for. Scribal sophistication in the ordinary sense of that term is not really compatible with the evidence, and there is no indication at all of authorial revision. Rather, the manuscripts present precisely the kinds of variants that might arise in a performer's recitation of the memorized text over a period of time. Without conscious thought of changing his text, he would have shifted tense, substituted synonyms, altered formulas, and restructured lines.[29] At some later date the performer who came to copy or recopy the memorized poem would be to a great extent like any other scribe. He would have attempted to reproduce his exemplar more or less faithfully, and like other scribes would have made his share of unconscious mechanical errors and exercised the same right of tampering with a reading here and there to improve or correct the written text before him. But he would differ from the ordinary scribe by having not only the written exemplar on the paper before him but also an aural memory of his own performances. He would, more or less consciously, have inserted his habitual expressions in preference to the readings of the exemplar. It would in turn serve as a check to his memory, a presence before him controlling what he wrote. It would halt the displacement or deletion of scenes and episodes. Some such process in the textual tradition of both manuscripts would account for both the demonstrable continuity in the copying of the manuscripts and for the kinds of consistent variation between the two texts. It is not possible to prove that this is what really happened, but it is less intrinsically improbable than the assumption that all of the variants not explicable by paleographic evidence are the result of ordinary scribal sophistication.

The question then arises, given the probability that both of the extant manuscripts are contaminated by the interposition of a performer's words, whether a critical edition is possible. The implications of Professor Baugh's studies are certainly discouraging. More recently, Professor William E. Holland's study of the variants in the manuscripts of Arthour and Merlin led him to the conclusion that in a verse romance transmitted through a series of minstrel copies the "attempt to delineate the descent of the manuscripts . . . must fail not merely because of an insufficiency of evidence (not enough manuscripts) but because such an attempt is based on an erroneous assumption about the nature of that descent" (p. 105). To this point, every study of the impact of oral delivery upon the texts of written manuscripts has indicated that it led inevitably to a rapid rate of textual corruption. The evidence from the manuscripts of The Wars of Alexander, however, suggests at least the probability that partial oral dissemination need not in fact have caused radical disruption of the text, that the highly formulaic diction of the poem must have been a stabilizing factor in the dissemination of the romance. Equally important for the modern editor, those formulas are extremely useful in establishing a critical text.

The steady and systematic use of evidence derived from the composition of a poem in formulas has not, to my knowledge, been a feature of earlier editions of Middle English texts.[30] For the Middle English alliterative revival, it is customary to argue, as John Finlayson does in his recent edition of the Morte Arthure, that the poets "would appear to have drawn on a common stock of alliterative formulas to express the events drawn from their source material, and in some cases to have created new formulas on the model of those with which they were familiar."[31] If this were

indeed the case, reliance upon formulaic evidence for decisions between variants would be fatuous, since a poet drawing mechanically upon a stock of formulas (as he would draw upon, say, a stock of Latin rhetorical topoi) would not be bound by the formula systems but free to change the conventional phrases in various ways to "make it new." However, the evidence overwhelmingly supports the opposite conclusion. The Wars of Alexander was not orally composed, at least not as an extemporaneous performance of the kind associated with the Yugoslav singers studied by Parry and Lord. But the author composed like an oral poet. Whether he also composed other poems orally is mere speculation, but that this poem is composed substantially in formulas and formula systems is ample evidence that he had a grammar of composition which enabled him to generate an almost endless number of phrases, clauses, and sentences which were also rhythmic half-lines of alliterative verse.[32] In short, the Wars-poet was able to do with pen and ink what generations of oral poets had done in extemporaneous performances. He did not just draw on a tradition of formulaic diction or borrow fixed formulas from other poems but wrote formulaically.[33]

The literate poet who composes formulaically is not precisely like the ordinary poet even when he is writing rather than singing his narrative. He composes from within a more limited sub-grammar of the language than the poet who does not compose in formulas, for that poet, though he may use tags and conventional phrases and may appeal as consistently to a listening instead of a reading audience, nevertheless has more of the syntactic and semantic resources of his language available to him. Professor Frederick Cassidy's report on research into the syntactic patterns utilized by the Old English poets who wrote the elegies and Beowulf reveals that only twenty-five basic syntactic patterns are used in those poems, considerably

fewer than were available to the writer of Old English prose.[34] I have not completed a grammar of The Wars of Alexander, but preliminary research reveals that a very few basic phrase structure rules will account for most of the off-line formulas.

I do not wish to imply that the sub-grammar utilized by the Wars-poet was in the least constricting or inadequate to his purpose, nor would it be appropriate to suggest that it is inflexible and absolutely regular. No natural language is. It was not constricting to the poet because it was learned by listening to and by reading a great deal of alliterative poetry, and it was not inadequate because it could generate an infinite number of grammatically and metrically acceptable lines. But from the view point of the modern student of his language, the options open to the formulaic poet are limited by that grammatical-metrical system, by the abstract syntactical, rhythmic, and (what is less well understood in spite of the emphasis upon fixed formulas) semantic collocations behind what he wrote.

The editor cannot assume that the poet composing formulaically achieved absolute regularity in his verse, though since he was able to reread and correct what he had written, his opportunities to achieve total regularity were much greater than for purely oral poets. Nor should one assume that he aimed at such regularity. None of the surviving Middle English alliterative poems is perfectly regular, especially in the alliterative patterns, though it is, I think, improbable that there is much conscious variation of formula systems to achieve the variety urged in rhetorical manuals. Apparent variety in the formulas, as R. F. Lawrence has recently argued, is much more likely to be motivated by the necessity to conserve basic rhythmic norms than by the desire to achieve artful innovation for reasons of ornamentation.[35]

It follows then that when the manuscripts vary, the reading that corresponds to an established formula system is likely to be original and that readings which are not consonant with the grammar of composition are probably corrupt. A few examples will demonstrate the uses of formulaic evidence in establishing the text.

1. At line 1129, the second half-line in A reads "as he þareon lokis." D differs only in the verb, which is the synonymous "wates." The readings are metrically and semantically interchangeable. Both lokis and wates appear elsewhere in the poem in the final stressed position, lokis at ll. 603, 677, 750*, 840*, 2942, etc., and wa(i)tis at ll. 131, 194, 265, 700, 781*, 956, 2930, 3630, 3636, 4129, 4776, etc. On positional grounds, there is no reason to choose between the two. However, the formula occurs again at ll. 956

and 3630, and in the first, both manuscripts agree in reading waites. Only A exists at l. 3630, but that it too reads waite is evidence for the probable originality of that word in this formula. Although the evidence is far too slight for certainty, it is also probable that loke was more common, even in the North, than waite by the middle years of the fifteenth century, when the A manuscript was written, and that waite is to be preferred as the more difficult reading.[36]

2. The principle of accepting the lectio difficilior, however, is a weaker reed in formulaic poetry than in verse less bound by semantic and syntactic conventions. At line 1738, the manuscripts differ again in the final stressed position. A reads "oure force to ministere," while D has "oure force to withstonde." Ministere "to control, govern" does not occur elsewhere in the poem. Indeed, using it in the sense of controlling a hostile force is an innovative extension of the normal meanings of the word (see OED, s. v., Minister, v. 7a). It makes excellent sense and, pronounced with the second syllable elided, is metrically regular. However, the D reading is clearly a manifestation of a common off-line formula system which occurs several times in the poem without substantial modification:

• . . . his wrothe to with-stand 2077
• . . . his force to with-stand 2357
• . . . ȝour faes to withstand 2804
• . . . oure force to withstand 3717
• . . . oure miȝtis to withstand 3766
• . . . ȝour saȝe to with-stande 5119

The D reading, which corresponds to the formula, is unlikely to have resulted from scribal error or sophistication; and the more difficult and, in some ways, aesthetically more satisfactory, reading of the A manuscript is to be rejected. If instead of composing formulaically the poet had been drawing on a tradition of formulas, quite the opposite decision would be required.

3. At l. 714, the A manuscript reads: "As be þe welken to wete quat suld come efter." D reads "Als be welkyn to wete what worth sall her-aftir." Henneman very sensibly suggested emendation of the line to "quat worth sall eftir," a course I followed in the first draft. However, there is some evidence that supports a more complicated emendation. As the following lines reveal, the half-line is a manifestation of a fairly common formula system:

• . . . quat suld worþe efter 171
• . . . he sall take eftire 360
• . . . how it be-tid efter 524
• . . . howe it worthis eftir 1718
• . . . þat sall fall eftir 1897A
• . . . þat falle sall after 1897D
• . . . how it sall tide eftir 2145
• . . . how it worthid eftir 2317 (probably spurious)

Both manuscripts agree in word order at ll. 1718 and 2145. Strictly speaking, the three instances which occur before the D fragment begins constitute only evidence for the practice of the A scribe and the existence of the formula system.[37] Recurrence of the D pattern "þat falle sall eftir" at l. 1897 is parallel to the reading here, and the two constitute additional evidence for the instrusion of habitual collocations of performers into that manuscript. It is, of course, not impossible that the poet himself employed both word-order patterns in this formula system, but the preponderance of evidence—the admittedly slight statistical evidence and the fact that the second D reading is of the avoided rhythmic pattern x/x/x—suggests that the line should read "quat suld worth eftir." That is, in this case, the A scribe had substituted the synonymous come for worth, and the D text had the performer's variant involving a shift in word order. The D text's her-eftir is probably motivated by the desire to meet the more common off-line rhythmic pattern of x/xx/x.

4. At line 1189, the A text reads in the off-line "he writhis him vnfaire," and D has "he wex wode wroth." The alliteration is suspect in D as being too heavy, but formulaic evidence provides the best reason for choosing the A reading:

• . . . angirs vnfaire 837
• . . . þat tened þam vnfaire 1212
• . . . deris þam vn-faire 2041
• . . . & turbled vnfaire 3637
• . . . was arȝed vnfaire 3873
• . . . berand vnfaire 3903
• . . . crabbid vnfaire 4566
• . . . & breed þaim vnfaire 4741
• . . . breis þaim vn-faire 4837

In spite of superficial differences, the lines cited above correspond to the rhythmic norm (x)/xx/ (perhaps x/xxx/ in l. 1212), and flexibility of syntactic relationships between the variant forms is made apparent in ll. 4741 and 4837. In this instance, the probability that the A reading is correct amounts almost to certainty.

5. At line 2199, the A manuscript reads in the off-line "þat citiȝens hatt," and D reads "þat citesyns er called." Neither reading is unsatisfactory in itself, but the preponderance of evidence suggests that neither is in fact correct. The formula recurs in the following lines:

• . . . & Anec was hatten 40
• . . . Pausanna was hatten 914A}
• . . . pausana he heght 914D}
• . . . my satroparis hatten 1913A}
• . . . my satropers halden 1913D}
• . . . Amont was hatten 2037A}
• . . . amonta was callyd 2037D}

---

285




• . . . Yssanna was hatten 2106
• . . . was Wyothy hatten 2150A }
• . . . was worthyly haldyn 2150D }
• . . . Platea was hatten 2297
• . . . Mocian was hatten 2540A }
• . . . of Mocian es haldyn 2540D }
• . . . þat angill is hatten 4720
• . . . was Marcipy hatten 5093

The evidence is clearly mixed. The A scribe with some consistency preferred the pattern ending in hatten, and the D scribe showed considerable uncertainty with variations on called, heght, and haldyn. However, agreement of both manuscripts at lines 2106 and 2297 suggests that the A text has retained the original form. In all probability, D is correct in having retained the syntactic pattern of the auxiliary plus past participle, A in having retained a form of hatten, and the corrected text should read "þat citizens er hatten."

6. Unfortunately, the D manuscript does not extend beyond l. 3467, but formulaic evidence may be used to correct the A text where it stands alone. For instance, at ll. 4775-4776, the text reads:

The text is not obviously corrupt as it stands, but there is a good deal of evidence that the second half of l. 4906 should read "& on þe trees waitis":

• . . . ay on þat dere waytis 265
• . . . quen he on him waites 956
• . . . as he þar-on lokis (wates, D) 1129
• . . . quen þai þare-on waite 3630
• . . . & on his ost waites 4129

7. The highly formulaic language sometimes led the scribes (or performers) to careless substitution of a common pattern for the correct reading. For instance, at l. 2829, the A manuscript reads "And etils to Ser Alexander eft to assaill." The D text reads "And ettlys þe sir Alexander efte to assayle." In this instance, citation of the immediate context will help to make the nature of the problem more clear; King Darius' mother Rodagoras writes to her son advising him not to continue his mad plan to attack Alexander:

The A reading corresponds to the common formulaic system "(and) + verb of motion, usually aires + (pronoun) + to Sire Alexsandire . . ." For line citations, see p. 0000 above. However, the attraction of that formula was a siren song for the A scribe, and the source reveals that D's þe is correct: "ut cum Alexandro iterum pugnaretis."[38]

8. The very best evidence for deciding between variants or for emending the single manuscript is the conjunction of formulaic evidence and the source. At lines 5359-5360, where only the A text survives, the text reads:

The alliteration is irregular in the off-line of 5360, for we would ordinarily expect alliteration to fall on the stressed syllable rad-. The formula system alliterates elsewhere on /m,d,s,a/, and we should expect the realization of the formula here to be "and sadly scho pleynes" on the model of the following lines:

• . . . & maynly hire pleynes 399
• . . . & depely þam playnt 1698
• . . . & sadly he pleyned 4204

Both the lack of regular alliteration and the existence of a formula system designed, as it were, to produce a regularly alliterating line on /s/ make the manuscript reading suspect. Fortunately, the formulaic evidence which suggests emendation to sadly finds corroboration in the source which reads: "Videns Candacis regina quod volebant se interficere filii sui, tristis effecta est nimis" (G1, fol. 159b).

9. Unhappily, the collection and analysis of formulaic evidence will not always permit the editor to decide between variants because of the complicating addition of performers' formulaic collocations to the texts. What at first sight appears to be a variant of the formula system just discussed appears in the off-verse of line 972. The A manuscript reads "augirly he wepis," and D reads "awgerdly pleynez." Alliteration is on the first syllable of augirly, and both readings are metrically regular. The D rhythm /xx/x is of more frequent occurrence than A's /xxx/x, but there is reason to regard both as variations on the same type. Comparison with the lines cited above suggests a generalization and a simple solution. There is obviously a formula system of the pattern "(and) + adverb + pronoun + pleynes," where the adverb and pleynes are stressed and where the particular adverb that appears is determined primarily by alliteration. The slot for the pronoun is unstressed and presumably could be filled with any personal pronoun of one syllable. Since pleynes is the stable element that defines the system in lexical terms, it would seem probable that D's pleynez is correct and the performer-scribe of the A manuscript was responsible for substituting the synonym wepis. Both manuscripts are thus corrupt, A having substituted the synonym and D having deleted the pronoun.

However, the D reading is itself regular metrically and seems to be based on another common formula system:

• . . . augirly granys 717
• . . . & augrily granys 3252
• . . . & bitterly wepis 963

The last cited form presents a thorny problem. It would seem to express the same "given essential idea" as "& augrily granys" or "awgerdly pleynez." Certainly, it has the same syntactic-rhythmic form of "(and) + adverb + verb" (x/xx/x). But if it is the product of the same formula system, the notion of lexical repetition as always a chief determinant of the formula system must be more limited than previous studies have suggested. The line of reasoning adopted earlier which would explain the A variant as a corrupted form of "augirly he pleynes" on the grounds that pleynes is the stable lexical item in that formula system is clearly inadequate. A new search of the poem for all instances of the off-verse pattern "(conjunction) + adverb + (pronoun) + verb" produced 233 lines, and among these were the following two lines which tend to corroborate the A reading:

• . . . baldly he wepis 5039
• . . . he rewfully wepid 2843A
• . . . rewfully he wepys 2843D

In short, there are at least three possible syntactic-rhythmic patterns (themselves variants of the same formula system?) which may be behind the two variants. Both readings are regular, and parallels for both are to be found elsewhere in the poem. One is certainly an error, a more or less conscious substitution for the original reading. Both may be corrupt, though it is equally possible that one of the two is authorial. In any case, there is no evidence to suggest that one reading is preferable to the other on any rational grounds.

If it is indeed true that this poem is composed formulaically and that there is the interposition of performers' variants in the history of both surviving manuscripts, the implications of such evidence are of considerable interest beyond the editorial conclusions I have drawn at length above. First, there is at least one transitional text between orally composed poetry and written poetry, written by a literate poet but composed within the rhythmic, grammatical, and lexical restraints of a grammar of composition probably oral in origin. There are almost certainly others. Secondly, modern studies of the use of formulas in medieval texts which survive in a single manuscript are possibly askew to the extent that the existence of systematic performers' variants are ignored or unknown. A study of formulas in The Wars of Alexander based on either manuscript alone will produce a quite distorted view of the poet's own practice, since in each instance, the poet's words are changed by the habitual formulaic collocations of the performer. It must not be forgotten either that both manuscripts are linked to an original exemplar by a continuous line of copying,

unlike the manuscripts of the metrical romances studied by Baugh and Holland. The appearance of performers' variants in manuscripts so descended suggests at least the possibility that other highly formulaic alliterative poems—and perhaps one should think of Old English as well as Middle English texts here—are similarly the product of collaboration between the poet and a line, long or short, of anonymous performer-scribes. It is, of course, possible that the manuscripts of the Wars are sports, that the process which left the two manuscripts to us was unique. Although the theory advanced in this article to explain the nature and number of variants is the most economical account of the facts, it may not be true. If it is, however, and if the manuscripts of the Wars are not unique, we need to examine again our theories of formulaic diction in the light of this new evidence.

Notes