Epistle to Cangrande Updated

Henry Ansgar Kelly (UCLA)

Dante Notes / September 28, 2018

The tragic incompatibility of Dante and the Epistle to Cangrande

In 1989 I published a book entitled Tragedy and Comedy from Dante to Pseudo-Dante,[1] intended as a first volume chronicling a history of different ways in which the word “tragedy” was understood in various cultures from Antiquity into medieval times. It was clear to me that the formulations of tragedy and comedy to be found in the Epistle to Cangrande are incompatible with Dante’s own views, especially as expressed in De vulgari eloquentia and throughout his Comedy, to the very end of Paradiso. For Dante, these genres had nothing to do with prosperity or disaster, or with lamentation or rejoicing or humor. Tragedy was a composition on a high subject in high style, whereas comedy accommodated multiple levels of subject and style.

I returned to Dante in my next book, Ideas and Forms of Tragedy from Aristotle to the Middle Ages, in which I contextualize Dante’s views and those of his commentators with other understandings of comedy and tragedy abroad in Italy and Europe then and earlier. I show not only the basis on which Dante considered Vergil’s Aeneid to be a tragedy, but also classified his own lyric poems, notably, Donne ch’avete intelletto d’amore, as tragedies.[2] A few years later, in Chaucerian Tragedy, I set out Boccaccio’s ideas of tragedy and comedy at great length, in contrast to Dante.[3]

My view is that Dante based his notions of tragedy and comedy on style and subject matter, not plot, whereas his commentators, who did not know the De vulgari eloquentia, had to rely on the plot criteria of the dictionaries, and it is inconceivable that Dante would have written what is contained in the Accessus section of Cangrande.

An early citation of Cangrande?

My best guess in the above studies was that Cangrande was compiled by an unknown person, whom I call Pseudo-Dante, in the last quarter of the fourteenth century, completed in time to be cited by Filippo Villani around 1400.

But recently there has been a new development. In 2003, Luca Azzetta established that in the early 1340s Andrea Lancia produced a commentary on the Comedy, finishing before the summer of 1343, in which he cited a letter by Dante to Cangrande on the division of Paradiso into two parts, prologue and execution: a clear reference to Cangrande 17.43 in the text as we now have it.[4] Azzetta and others following him conclude that Lancia possessed the whole letter, which, in my terms, consists of the Dedication, the Accessus, and the Exposition, with various linking parts.[5]

Robert Hollander, the chief adversary of my original position,[6] believes that the discovery of the Lancia citation has collapsed my whole “house of cards” and consigned it to the dustbin of history.[7] Azzetta concurs in this judgment, though not so forcefully, but adds that Hollander had already refuted my case in his 1993 book.[8] Azzetta does not at this point refer to the Cangrande Dispute carried on between me and Hollander in 1994 in the pages of Lectura Dantis,[9] the result of which might be termed a “standoff” rather than a victory on one side or the other.

Azzetta’s conclusion that Lancia had the whole Epistle to Cangrande in his hands–and I acknowledge that he may well be right—does not necessarily alter my assessment of how the compilation was put together, but only when it was done. As Zygmunt Barański says, “There is nothing in Lancia’s statement to prevent one from concluding that the Epistle is an early rather than late Trecento forgery, since the commentator’s recourse to the letter does nothing to resolve the many technical and ideological problems associated with it.”[10]

If Azzetta is right, that Lancia saw the entire Cangrande, whoever compiled it (the Compiler) would have had to finish his patchwork by 1340 or so. I posit that he would have possessed a genuine letter by Dante of some three paragraphs, dedicating Paradiso to Cangrande, and linked this letter to his own (or someone else’s) Exposition of the first few lines of Paradiso, prefaced by an Accessus. For the latter, he modified an already existing work that I call the Proto-Accessus, based largely on Guido da Pisa. Whereas the Proto-Accessus explained the whole Comedìa, the Compiler changed it to emphasize Paradiso in particular.

Alberto Casadei has suggested Verona as a likely place for the creation of Cangrande. Not only was Verona a place where one might expect to find such a letter, but also the city possessed an important manuscript of Pliny’s Naturalis historia, which might have provided a suggestion and a model for it: it contains the author’s dedication of the work to the future emperor Titus.[11]

Even granted that Lancia saw Cangrande as it now exists, it cannot be demonstrated that any other commentator had access to the intact composition before it was cited by Villani.

In my new reconstruction (based on the assumption that the whole Cangrande was created around 1340), the Proto-Accessus would have continued to circulate independently. It was seen by Boccaccio, who felt free to reject or modify its positions—something he would not have done if he had known that these ideas were those of Dante himself.[12] Azzetta for his part agrees that Boccaccio and others did not see the whole of Cangrande; he argues rather that the Accessus and Exposition sections somehow got detached from the Dedication, and that it was only the detached Accessus that some commentators saw, while others saw only the detached Exposition.[13]

Furthermore, Azzetta believes, this dismemberment of Cangrande happened very early, a few years after Dante’s death, in time for the Accessus to be used by Jacopo della Lana sometime between 1324 and 1328.[14] For my part, however, I do not believe that Lana’s text shows any close similarity to the Accessus, except perhaps for some phrases that he probably borrowed from an early version of Guido da Pisa’s commentary,[15] and I argue at length that the Proto-Accessus is dependent on Guido, not vice versa.[16]

Azzetta believes that my treatment of the latter question, the priority of Guido, was already refuted not only by Hollander but also by Thomas Ricklin.[17] He instead addresses Casadei’s case for Guido’s priority.[18] But his arguments that Guido drew on the Accessus[19] show at most that it is possible, not probable; the influence could be in the other direction. I hold that the latter view is more plausible: that is, it seems more likely that the Proto-Accessor cut down on Guido’s wordiness, than that Guido expanded on the terseness of the Cangrande Accessus. Take, for instance, the two quotations from the same passage of Aristotle’s Metaphysics in two different Latin translations. Ricklin has shown that Dante (as author of Cangrande) could have found both forms in the commentary of Albert the Great;[20] however, though we know that Dante was familiar with Aquinas’s commentary on the Metaphysics, it has not been established that he made use of Albert’s. It seems just as likely an explanation that Guido’s use of the Metaphysics inspired the Proto-Accessor to consult Aristotle’s work (in a different Latin translation) and quote further from it.

To return to Andrea Lancia: Azzetta does not accept the identification of Lancia as the author of the Ottimo Commento in any of its versions (whether 0 Ottimo, 1 Ottimo, or 2 Ottimo in my terminology),[21] because the passages from Dante’s Convivio cited in 2 Ottimo are different from the passages cited in Lancia’s Chiose.[22] He identifies the author of 3 Ottimo as “Amico dell’Ottimo,” the oldest manuscript of whose work, Morgan 676, was copied by Lancia sometime between 1345 and 1355.[23] But while the 3 Ottimo text may show knowledge of the Cangrande Exposition, there is nothing there to indicate that the author knew the Accessus.[24]

The possibility is raised by these commentaries, 3 Ottimo and Lancia’s Chiose, that what Lancia saw and cited was not the whole Epistle to Cangrande but only a preliminary form consisting of the Dedication and the Exposition, whether the Exposition was composed by Dante himself or by someone else (the Expositor), and joined to the Dedication by the Expositor or by another person (the Compiler). If so, then the same Compiler later on, or else a Deutero-Compiler, would have taken the Proto-Accessus and modified it into the Accessus by emphasizing Paradiso, and placed it between the Dedication and the Exposition to form the Cangrande cited by Villani, that is, the whole Cangrande that we know today.

The Compiler’s Bad Beginning and Bad Ending: Cursus Problems

I have no objection on the basis of contents to acknowledging Dante as the author of the Exposition as well as of the Dedication. Casadei has put forth an argument that it betrays a non-Dantean understanding of gloria,[25] a point plausibly contested by Saverio Bellomo.[26]

However, I do have two stylistic objections, one vincible and the other invincible. The former is the unlikelihood that Dante, as a champion of the illustrious vernacular, would translate the lines of his vernacular masterwork into Latin before commenting on them. My insuperable objection is that Cangrande does not match Dante’s cursal style, whether at the finish of the epistolary Dedication or at the end of the Exposition, when the epistolary mode resumes.

I have demonstrated that there is a notable breakdown at the end of the Dedication (4.12-13), not only in the text as established by Giorgio Brugnoli and others,[27] but also as emended by Enzo Cecchini.[28]    

Azzetta follows Cecchini’s corrections, except for adding one of his own, changing the unsonorous conferri videri potest in 4.12 to conferri potest videri, a first-class planus.[29] He claims, without offering justification, that Dante followed the style of the papal curia for his periods, and the style of the Orleans school for internal pauses.[30]

He gives a table of clausal endings for the first four paragraphs, ending with 4.13.[31] The first four clauses of 4.13, which he pronounces regular, are as follows:

videbar expressisse      trispondaicus

proposito fui[t]            planus [N.B. insipidus]

gratie vestre                planus [N.B. insipidus]

vitam parvipendens     trispondaicus

The trouble is, Dante did not consider trispondaicus forms to be acceptable (unlike later writers);[32] and the planus type that appears here is the enclitic form that Dante disparaged in the De vulgari eloquentia (2.6.4-5) as insipidus, giving the example of dominam Bertam.[33] Moreover, the first planus insipidus here was created by Cecchini, who rejected the usually accepted quod de proposito for the reading quod de proposito fui, changing fui (which makes no sense) to fuit.

Cangrande clearly states that Dante is sending the whole of Paradiso to Cangrande,[34] but if Casadei is right to argue that Dante could not have sent the finished cantica to Cangrande while Cangrande was still victorious,[35] and if Dante really wrote the Dedication, the Compiler must have altered the language of 3.11 to refer to the whole sublimem canticam. As Casadei notes, this phrase is not cursal,[36] and neither is the penultimate clause, non offero.

Critics of “cursal attacks” on Cangrande constantly point out that the strict cursus is not necessarily followed in non-epistolary prose. However, even though cursal experts basically agree with this proposition, there are important distinctions to be made. Both Paget Toynbee and Peter Dronke argue that authors who were used to observing the rules of cursus in their epistles would tend more often than not to follow the same conventions in other modes of writing. Dante was no exception; he has a high percentage of cursal endings in the De vulgari eloquentia and Monarchia.[37] Of the De vulgari eloquentia Dronke says, “A distinctive pattern of cadences can be seen, which is so startlingly different from that of the fitful cadences in the exposition to Cangrande that it is very difficult indeed to imagine that both could stem from the same author.”[38]

Dronke studied only periods, but I extend the analysis to all pauses (interior phrases and clauses) and show that cursal observance is especially high at the beginning and end of non-epistolary works. Of the 394 clauses of the Accessus and Exposition, only 38% conform to Dantean cursal forms, whereas there is a 60% rate in the 387 clauses of De vulgari eloquentia 2-10.[39] See Table 1 for the cursus-endings favored by Dante.

De vulgari eloquentia was not finished, but Monarchia, a treatise undoubtedly by Dante, has a fairly regular ending, with 47 out of 68 cadences strictly cursal, a 69% rate, unlike the very straggly ending of Cangrande, which has a mere 39% cursal conformity (discussed below). The ending of the disputed Questio de aqua et terra, the whole of which is in the form of a public epistle, does not fare much better, at 40% (see Table 2).[40] True, Questio gets off to a regular start, but then falls apart to give it a 60% rate, with 18 regular clauses out of 30. Monarchia, in contrast, comes in at 87%, with 26 Dantean forms out of 30 (Table 3).

If the whole Epistle to Cangrande were by Dante, there would be reason to expect that the end would be entirely regular, like the beginning; for, once the Exposition has concluded, the text reverts to addressing Cangrande. That is, it falls again into the epistolary mode. In the first section, 32.88, there are six clauses, three of them non-Dantean, including a planus insipidus (Magnificentia vestra):[41]

            prologi in generali N

exponam ad presens P2

familiaris angustia T1

derelinquere oporteat N

Magnificentia vestra N(P3)

expositionem facultas P1

In the last six sentences of Cangrande, in 32.88 and 33.89-90, I count 23 clauses, only 9 of them Dantean, a rate of only 39% (Table 2). The ending of the penultimate sentence, utilitatem et delectationem, is a real clunker, and it is followed by other unacceptable cadences. Here are the last 9 clauses, showing a string of 6 non-Dantean forms in a row:

utilitatem et delectationem N   (nil)

principio seu primo N(S5)          trispondaicus

videlicet Deo N(P3)                    planus insipidus

ulterius queratur N(S4)               trispondaicus

A[lfa] et O[mega] N(S2)            trispondaicus

principium et finis N(S5)            trispondaicus

Iohannis designat P1                 planus 

terminatur tractatus P1              planus

secula seculorum V1                 velox

The use of the planus insipidus and trispondaics that we find here is typical in the prose style of the mid-fourteenth century, as in Petrarch’s usage, but not in Dante’s time.[42]

Azzetta is content that the final sentence ends in a velox[43] (preceded, we note, by two solid planus forms). But surely a stylist as precise as Dante would not advance to the concluding statement in an important letter in such an inelegant way.


The discovery made by Luca Azzetta of Andrea Lancia’s use of the Epistle to Cangrande around 1343, whether in the form in which we know it, or with the Accessus missing, is extremely important and forces us to revise our ideas about it. If one believes, as Azzetta does, that Lancia saw the whole letter and that Dante was responsible for it all, one must speculate about how and why and when it was almost immediately dismantled, with the Accessus and Exposition parts circulating separately, and the Dedication circulating not at all. Or, if one believes that Dante’s hand is only in the Dedication, as I do, one must try to judge when and how the other parts were added.

The simplest hypothesis, from my point of view, is that someone pretending to be Dante (whom we can call Pseudo-Dante), around 1340 took a genuine letter of Dante’s addressed to Cangrande and added to it an Exposition of the first lines of Paradiso, which was then seen by Lancia. The Accessus section, which was based on a hypothetical Proto-Accessus (abbreviating Guido da Pisa), may also have been added at this time. If so, the Proto-Accessus must have continued to circulate by itself. If, however, the Accessus was not in the version of the letter seen by Lancia, it must have been added later by another expander of the Dantean letter, who modified the Proto-Accessus to put particular emphasis on Paradiso.

In short, the expander or expanders of Dante’s original lacked familiarity not only with Dante’s cursal style but also with his ideas of tragedy and comedy elaborated in De vulgari eloquentia. Unlike some other Dantean contradictions, this one cannot be explained by saying that the poet changed his mind later on. Actually, he did change his mind on one point. When he wrote the Vita nova, he was not thinking about tragedy, and he only later reanalyzed his lyric poems as tragic. But the same is not true about changing from a virtue-based idea of tragedy to a disaster-based notion. When Dante says toward the end of Paradiso he was defeated more than any comico or tragedo in attempting to describe and praise the beauty of Beatrice (Par. 30.22-24), this is hardly consonant with the pronouncement of Cangrande that the matter of tragedy proceeds from what is admirable and quiet to what is filthy and horrible, and that comedy is the reverse.[44] No, this is not Dante talking.

[1]H. A. Kelly, Tragedy and Comedy from Dante to Pseudo-Dante [TC] (Berkeley, 1989, repr. Eugene OR: Wipf and Stock, 2004). Chapter 2, “The Chronology of the Proto-Accessus,” pp. 11-18, appeared in an earlier form: “Dating the Accessus Section of the Pseudo-Dantean Epistle to Cangrande,” Lectura Dantis [Virginia] no. 2 (Spring 1988) 93-102, link:


[2]Ideas and Forms of Tragedy from Aristotle to the Middle Ages (Cambridge, 1993, corr. paperback ed. 2005), pp. 144-57: “Dante and His Commentators.” For a briefer account, see my “Interpretation of Genres and by Genres in Medieval Literature,” in Interpretations: Medieval and Modern, ed. Piero Boitani and Anna Torti, J. A. W. Bennett Lectures, no. 7: Perugia, 6-8 April 1992 (Woodbridge, 1993), pp. 107-122, esp. 116-18. Recently, Thomas Ricklin, “Indagine su un disguido epistolare: l’Epistola a Cangrande fra Verona e Padova,” Medieval Letters: Between Fiction and Document, ed. Christian Høgel and Elisabetta Bartoli (Turnhout, 2015), pp. 369-79, discusses the possibility that Dante in the Cangrande Accessus was reacting to Mussato’s Ecerinis; but he does not take note of how both Mussato and Dante fit into discussions of tragedy. I conclude in Ideas and Forms, p. 144, that “the ferment over Seneca in Padua seems not to have attracted the attention of Italy’s greatest poet.” For my account of Mussato and his circle, see Ideas and Forms, pp. 134-43 (“Seneca at Padua”); “Interpretations of Genres,” p. 116.

[3]Chaucerian Tragedy (Cambridge, 1997, paperback ed. 2000), pp. 11-38: “Background: Boccaccio’s Non-Tragedies.” See also pp. 6, 39, 80-81.

[4]Luca Azzetta, “Le chiose alla Commedia di Andrea Lancia, l’Epistola a Cangrande, e altre questioni dantesche,” L’Alighieri 21 (2003) 5-76, esp. pp. 23, 38. See also his editions of Lancia and Cangrande: Andrea Lancia, Chiose alla “Commedia,” ed. Luca Azzetta, 2 vols. (Rome, 2012); Epistola XIII, ed. Luca Azzetta, in Dante Alighieri, Le Opere (Nuova Edizione Commentata), gen. ed. Enrico Malato et al., vol. 5: Epistole; Egloghe; Questio de aqua et terra, ed. Marco Baglio et al. (Rome, 2016), pp. 271-487: introduction (pp. 273-324); text and commentary (pp. 326-417); appendix showing the letter’s “indirect tradition” (pp. 418-87).

[5]See my edition of the letter, TC 102-11, with cursus markings of the clauses. Thomas Ricklin, “L’Epistola a Cangrande: Elementi assodati e nuovi spunti a proposito di un testo contestato,” Rivista di Studi Danteschi 15 (2015) 66-97, at 89-95, though believing that Dante wrote the whole of Cangrande, suggests that Lancia saw and cited only an excerpt from it, at most an early anonymous accessus to Paradiso based mainly on the Cangrande Accessus.

[6]Robert Hollander, Dante’s Epistle to Cangrande (Ann Arbor, 1993).

[7]Robert Hollander, “Due recenti contributi al dibattito sull’autenticità dell’Epistola a Cangrande,” Letteratura Italiana Antica 10 (2009) 541-52, esp. 543, 548. Hollander infers that my “unusual silence” in the matter indicates that I am at a loss for words. The truth is that, because I have been toiling in other academic fields, I had not noticed Azzetta’s discovery until recently.

[8]Azzetta, Epistola XIII p. 429.

[9] The Cangrande Dispute [CD], in Lectura Dantis (Virginia) nos. 14-15 (1994) 61-115: Henry Ansgar Kelly, “Cangrande and the Ortho-Dantists,” pp. 61-95; Robert Hollander, “Response,” pp. 96-110; Kelly, “Reply,” pp. 111-15.

[10]Zygmunt G. Barański, “The Epistle to Can Grande,” in The Cambridge History of Literary Criticism, vol. 2, The Middle Ages, ed. Alastair Minnis and Ian Johnson (Cambridge, 2009), pp. 583-89, at 589.

[11]Alberto Casadei, “Sull’autenticità dell’Epistola a Cangrande,” Ortodossia ed eterodossia in Dante Alighieri; Atti del convegno di Madrid (5-7 novembre 2012), ed. Carlota Cattermore et al. (Madrid, 2014), pp. 803-30, esp. 823-25; also, idem, “Essential Issues Concerning the Epistle to Cangrande,” Medieval Letters, ed. Høgel and Bartoli (2015), pp. 381-92, esp. 391-92; his suggestion is summarized by Azzetta, Ep. XIII, p. 429.

[12] TC 47.

[13]Azzetta, Ep. XIII, pp. 418-20, 423, 445; cf. TC 17.

[14]Azzetta, Ep. XIII, p. 423.

[15] TC 23-24.

[16]TC 12-17, 19-23.

[17]Azzetta, Ep. XIII, p. 429, citing Thomas Ricklin, ed., Dante Alighieri, Das Schreiben an Cangrande della Scala (Hamburg, 1993), pp. xliv-xlix.

[18]Azzetta, Ep. XIII, pp. 429-32. See Casadei, “Sull’autenticità,” pp. 811-15; “Essential Issues,” pp. 390-91. See also Casadei’s essay, “Il titolo della Commedia e l’Epistola a Cangrande, Allegoria 60 (2009) 167-81, rev. in idem, Dante oltre la “Commedia” (Bologna, 2013), pp. 15-45.

[19]Azzetta, Ep. XIII, pp. 429-32.  

[20]Cf. Azzetta, Ep. XIII, pp. 343-44.

[21]TC 25.

[22]Azzetta, “Le chiose,” pp. 31-32. See further Azzetta, “La tradizione del Convivio negli antichi commenti alla Commedia: Andrea Lancia, l’Ottimo commento, e Pietro Alighieri,” Studi Danteschi 5 (2005) 3-34; idem, “Tra i più antichi lettori del Convivio: Ser Alberto della Piagentina, notaio e cultore di Dante,” Rivista di Studi Danteschi 9 (2009) 57-91.

[23]Azzetta, Ep. XIII, pp. 444-46. See Azzetta, “Andrea Lancia, copista dell’Ottimo commento: Il MS New York, Pierpont Morgan Library M 676,” Rivista di Studi Danteschi 10 (2010) 173-88.

[24]Azzetta, Ep. XIII, pp. 445, 447. In TC 38-39, I discussed the possibility that 3 Ottimo may have drawn on the Cangrande Exposition (or on a Proto-Exposition); see also CD 114, replying to Hollander’s response, pp. 109-10.

[25]Casadei, “Sull’autenticità,” pp. 815-25; “Il titolo” (Dante oltre, 2013), pp. 34-43.

[26]Saverio Bellomo, “L’Epistola a Cangrande, dantesca per intero: ‘A rischio di procurarci un dispiacere,’” L’Alighieri 45 (2015) 5-19, , pp. 12-15. Casadei holds that Dante’s meaning is that God’s glory “penetrates through the universe,” on the one hand, and, on the other hand, “shines more in one place than in another,” whereas the Exposition takes the verbs to be synonymous, a hendiadys: God’s glory “penetrates and shines through the universe more in one place than another.” Bellomo argues that the latter interpretation is not only possible but may well be the authentic one. Casadei responds in “Sempre contro l’autenticità dell’Epistola a Cangrande,” Studi Danteschi 81 (2016) 215-45, pp. 334-46, Bellomo’s points are also countered by M. Signori, “Sulla distinzione di luce et Gloria nel ‘Paradiso’ dantesco,” Italianistica 45 (2016) 51-66.

[27] TC 62-68, 103-04.

[28]CD 113; I draw here on Enzo Cecchini, “Testo e interpretazione di passi dell’epistola a Cangrande,” Res publica litterarum 15 (1992) 115-29 at p. 117, carried over in his later edition: Dante Alighieri, Epistola a Cangrande, ed. Enzo Cecchini (Florence, 1995), p. 6

[29]Azzetta, Ep. XIII, pp. 306, 340.

[30]Ibid., p. 319. It may be that Azzetta was misled by my tables in TC 83-99, where I contrast “periods” with “clauses”; by the latter I mean “all clauses, including periods,” but I see that I could be taken to mean “internal clauses only.”

[31]Azzetta, Ep. XIII, p. 320.

[32]TC 80; CD 66.

[33]See TC 79-82; CD 66-67.

[34]As Azzetta, Ep. XIII, pp. 285-86, admits.

[35]See Casadei, “Essential Issues,” pp. 382-84, and earlier writings cited there.

[36]Casadei, “Sull’autenticità,” pp. 826-27 n. 3.

[37]Paget Toynbee, “The Bearing of the Cursus on the Text of Dante’s De vulgari eloquentia,” Proceedings of the British Academy 10 (1921-23) 359-77; Peter Dronke, Dante and Medieval Latin Traditions (Cambridge, 1986), pp. 103-11; cf. TC 63-64; CD 66. 

[38]Dronke, Dante, pp. 105-06.

[39]TC 88, 95.

[40]CD 74, and Appendix 2 B, pp. 91-92.

[41]TC 111.

[42]TC 66-67.

[43]Azzetta, Ep. XIII, p. 279. I must note here that in my edition of Cangrande, I labeled the final clause wrongly, putting t1 rather than v1 (TC 111, repeated in CD 71, 73); but the mistake did not affect my calculations (that is, I rightly counted the cadence as a velox rather than a tardus).

[44]Kelly, Ideas and Forms of Tragedy, pp. 148-49; CD 76.