 | |  |
Notes
Compare for instance the introductions to the following works: Albert Henry, Les Oeuvres d'Adenet le Roi, Vol. I, "Biographie d'Adenet," "La tradition manuscrite," I (Brugge, 1951); Jean Rychner, ed. Le Lai de Lanval, by Marie de France (Genève-Paris, 1958); William Roach, ed. The Continuations of the Old French Perceval of Chrètien de Troyes, Vol. I, "The First Continuation" (1949).
A rather recent survey of the different works concerned with textual criticism is offered by Edward B. Ham in his article, "Textual Criticism and Common Sense," RPH, XII (1959), 198-215. Vinton A. Dearing, A Manual of Textual Analysis (1959), is predominately concerned with the theoretical aspect of the problem and refers in his introduction to a number of other theoretical studies.
It is interesting to note, in this connection, that Paul Maas, after more than thirty years of practical experience which lay between the first and third edition of his Textual Criticism (1958), judged it appropriate to restrict the notion of peculiar error to the more limited notion of indicative error and even more restrictively to notions of separative and conjunctive errors (op. cit., supplement).
Comp. for instance: J. Andrieu, "Principes et recherches de critique textuelle," Mémorial des Études Latines (Paris, 1943), pp. 458-474.
A. C. Clark, Descent of Manuscripts (1918); Louis Havet, Manuel de Critique verbale appliquée aux textes latins (Paris, 1911), Chapter LXXXI, "La classification généalogique des manuscrits." Also: F. W. Hall, A companion to Classical Texts (1913), Chapter "Recension."
This trend in classical philology thereupon permitted such synthetic works as Paul Collomp's theoretical revaluation of the different methods of reconstructing the stemma (op. cit.), Giorgio Pasquali's, Storia della editione critica del testo, 2nd ed. (Firenze, 1952), and A. Dain's exhaustive collection of the information which is available to the critic through the study of manuscripts: Les Manuscrits (Paris, 1949).
Compare for instance the vehement rejection by A. Dain, op. cit., pp. 154-171. It is, however, interesting to note that the same views were proclaimed more than a half century ago by the distinctly antitraditional critic A. E. Housman, for example in his ed. of M. Manilii Astronomicon, I (London, 1903), pp. xxx-Liii passim.
Alexandre Micha, ed. Cligès (Paris, 1957); Bateman Edwards, A Classification of the Manuscripts of Gui de Cambrais "Vengement Alixandre" (1926); Albert Henry, op. cit. in note 1.
Compare for instance William Roach, ed. Le Roman de Perceval ou le Conte du Graal, by Chrétien de Troyes (Genève-Lille, 1956), p. xi: ". . . la critique textuelle des oeuvres littéraires du moyen âge a été une longe suite d'hypothèses instables et de décisions arbitraires des éditeurs, régulièrement abandonnées à mesure qu'avancent nos connaissances dans le domaine de l'ancienne langue. La difficulté réside dans le fait qu'il faut choisir entre le risque de laisser subsister une simple erreur de copiste et celui d'écarter une forme que l'auteur et ses contemporains auraient jugée parfaitement correcte et naturelle."
Mario Roques, ed. Erec et Enide (Paris, 1952), pp. iv and vi: 'Il est . . . souhaitable que le texte des ouvrages de Chrétien soit publié, si possible—sans préjudice des reconstructions méthodiques qui pourront encore en être tentées—d'après un manuscrit . . . Nous nous proposons d'en donner la reproduction intégrale pour servir à la lecture des romans de Chrétien dans une forme authentiquement médiévale . . ."
Alexandre Micha, La tradition manuscrite des romans de Chrétien de Troyes (Paris, 1939), p. 67.
Compare for example, Konrad Zwierzina, "Die Innsbrucker Ferdinandenhandschrift kleiner mhd. Geichte," in Festgabe für Samuel Singer (Tübingen, 1930): "Aber ich möchte warnen davor, . . . dass man sich überhaupt um die Aufstellung eines Stemmas und die dem Stemma entspringenden Überlegungen nicht viel kümmere. Denn nur in Verbindung mit der Beachtung des sich systematischer Zusammenstellung ergebenden Stammbaums kann die Beachtung der Eigenart der Einzelhs. fruchtbar werden" (p. 148). Compare also by the same author: "Überlieferung und Kritik von Hartmanns Gregorius," ZDAL, 37 (1938), 129-216 and 356-416; "Die Kalocsaer Handscrift," in Festschrift Max H. Jellinek (Wien-Leipzig, 1928), 206-232.
Joseph Bédier, "La tradition manuscrite de Lai de l'Ombre," Rom, LIV (1928), 161-196 and 321-356.
The only appreciation which acknowledges, although with many restrictions, the merits of this concise study seems to be the review by F. M. Salter [RES, XIII (1937), 341-352]. Current objections to Greg's methods are that their practical results are as limited as those of any other method, which indeed was acknowledged by Greg himself (Calculus, 53 seq.). The reasons for the misunderstanding of Greg's Calculus may have been the title of the work and its unusual symbols and formulas. The study, however, has no connection with mathematics; its basis is wholly logic, and very sound logic, hardly expressible by means other than symbols.
Dom Henri Quentin, Essais de critique textuelle (Ecdotique) (Paris, 1926); Hermann Paul, "Zur Nibelungenfrage," BGDSL, III (1876), 337-490, was first to apply probability calculus in attempting to answer a number of questions pertaining to versification and rimes. Paul's study escaped my attention and therefore I attributed in GL, V, 3, Supplement 1962, p. 78, the priority to Jean Fourquet, "Le paradoxe de Bédier," in Mélanges 1945, II, Etudes littéraires, Publications de la Faculté des Lettres de l'Université de Strasbourg (1946), 1-16; F. Whitehead and W. Pickford, "The Two-branch Stemma," BBSIA, No. 3 (Paris, 1951), 83-90; Jean Fourquet, "Fautes communes ou innovations communes," Rom, LXX (1948-49), 85-95; Rev. by Mario Roques in Rom, LXIX (1946-47), 116-118. Arrigo Castellani, Bédier avait-il raison? Discours Universitaires, Nouvelle Série, XX, Fribourg (Suisse): Editions Universitaires, 1957.
In his article in Mélanges, Fourquet attempted to prove with probability calculus that a dichotomous stemma is more plausible than a three-branch stemma. Later, persuaded by the reasonings of Whitehead and Pickford, that there had been fallacy in his demonstration, Fourquet revoked—according to the testimony of H. Niewöhner ["Kunst oder Methode," BGDSL, LXXIX (Halle, 1957), 422]—his original thesis. In response to Mario Roque's review, Fourquet attacked the problem in a different manner, concluding that in a series of cases in which the witnesses do not offer sufficient evidence, the stemma would remain undetermined; secondly, that in many cases, and under seriously limiting conditions, a dichotomous stemma can be proved; and finally, that a three branch stemma can never be proved. Our demonstration will show quite the opposite of Fourquet's conclusion. However, the importance of Fourquet's studies lies in their innovation in method for he attempts for the first time to evaluate the plausibility of a hypothetical solution by probability calculus.
For a detailed analysis of the problem comp. V. A. Dearing, A Manual of Textual Analysis (1959), p. 18 seq.
We shall use from now on the symbol "A" for the archetype, the symbols X and Y for inferential manuscripts and A, B, C for extant manuscripts.
Archibald A. Hill, "Some Postulates for Distributional Study of Texts," Studies in Bibliography, III (1950-51), calls this approach distributional study, and distinguishes it from genealogical, external and literary study. The following analysis intends to demonstrate that an adequate interpretation of the distributional data permits us to establish a unique genealogical stemma. In this decisive point, our stemmata differ from the trees constructed by A. Hill, since the latter ones, admittedly, are non-unique and the author does not claim to describe by them the genealogical and "historical" relationship of the manuscripts.
The mathematical part of the proposed method has been designed with the constant help of Professor Roger Pinkham of Rutgers University, later reexamined by Professor Ronald Pyke of the University of Washington. I wish to express my thanks to both of them for the help they have given me. Nevertheless, I myself must assume full responsibility for any errors which there may be.
It is, of course, impossible to evaluate the number of potential changes contained in any particular exemplar. In practice, however, we can overcome this difficulty by the proper definition of the potential changes. In our calculations, we shall for instance assume that the number of potential changes contained in an exemplar is equal to the number of words of its text. We thereupon postulate that a scribe copying an exemplar can introduce one, but only one, change per word. As a consequence of our postulate, we must count one difference per word even when the comparison shows two or more differences in a particular word of the text; also, we must disregard the transpositions, omissions and additions of verses, because it is impossible to estimate their potential number. Since the omissions and additions of different states are usually very indicative for the genealogical relationship of the texts, we must expect that a properly established stemma will not contradict the evidence contained in these indices. The definition of a difference is a difficult problem which cannot be discussed here. Generally one may say, however, that the definition itself matters less as long as it is one which can be applied consistently.
The equations of the following table are based on a "reasonable" estimate of the true probability; in order to estimate what values of the unknown would make the observed quantities as probable as possible, it would be necessary to introduce the principle in statistics called maximum likelihood. The resulting equations are rather complex and can be solved only by an adequate computer. In the following, we shall work with the "reasonable" estimates of the probabilities because we had no machine at our disposal. The unmaximized equations which we shall use have the advantage of being easily controllable by a non-mathematician. Compare Roger Pinkham's explanation of the principle of maximum likelihood in my report, "Statistical Methods in Textual Criticism," General Linguistics, Vol. 5, no. 3 (Supplement, 1962), p. 83 seq.
Our statistics are based on the text as established by Joseph Bédier, ed. Le Lai de l'Ombre par Jean Renart (Paris, 1913), and John Orr, ed. Jehan Renart, Le Lai de l'Ombre (Edinburgh, 1948).
The fact that some texts or parts of them may be the product of memorial transmission affects the texts in a similar way as contamination. The fact that other texts again are transcripts of dictation, affects, to be sure, the character and the extent of the changes introduced by the scribes, but evidently has no bearing on those factors which constitute the genealogical relationship of the witnesses nor on the possibility of applying the probability calculus. Finally, the proof that the one or the other line of descent may be based on an exemplar reedited by the author himself escapes in any case textual criticism in the strict sense of the word.
The table also indicates that, in our case, the archetype contained a very small number of recognizable errors, it shows namely that the important deviation from one (.034 and .035) is not produced by a corrupted archetype, since it appears only in the groups ADF and BDF. It must therefore be caused by a factor which is common to DF only; the archetype, being common to all four manuscripts, would necessarily cause a similar deviation in all four groups.
| | |