Applying Bayesian Phylogenetic Methods to the Stemmatics of the Divine Comedy

Presentation at the MPI-SHH WIP, December 2019

Transcript

1. Applying Bayesian Phylogenetic Methods to the Stemmatics of the Divine Comedy Tiago Tresoldi CALC/DLCE/MPI-SHH Jena, 2019-12-05
2. Stemmatics • "Copies introduce changes" • Developed along with religious and legal exegesis (e.g. "witnesses") • Mutation traditionally seen as bad ◦ "truth" of the archetype ◦ "changes" are "errors" to be removed ◦ "purity" and "contamination" • Scientific discipline developed in the same cultural framework of evolutionary biology and historical linguistics ◦ Karl Lachmann (1793-1851) ◦ Joseph Bédier (1864-1938) • Adoption of phylogenetics has been slow ◦ "New stemmatics" (Bordalejo 2006; Robinson 2016) ◦ "éloge de la variante" (Cerquiglini 1989) First actual stemma (Schlyter and Collins, 1827)
3. Learn stemmatics in one minute! Scheme of descent of the manuscripts of Pseudo-Apuleius' Herbarius by Henry E. Sigerist (1927) (Source: "Textual Criticism", Wikipedia)
4. Text in evolution • Easy to find analogies with biology and linguistics (e.g. admixture) • Copying manuscripts is hard ◦ Copyists didn't even necessarily understand what they were writing! • "Agency" of changes and of reconstructions • Manuscripts are not static • Changes are not necessarily independent • Hard polytomies are real, common and expected "Christine de Pizan in her study" in Cent Ballades. British Library, Ms. Harley 4431, f° 4.
5. Divine Comedy, manuscript Triv Divine Comedy, manuscript Ash
6. Notes, glosses and corrections by Pietro Alighieri (c. 1350)
7. Divine Comedy • One of the most successful medieval works ◦ More than 600 complete manuscripts ◦ Canterbury Tales at 80 (inc. incomplete ones) • Written between 1306-1321, circulating before conclusion • Read and copied by speakers of different varieties • "Tradition" characterized by intentional "contamination" ◦ The case of Boccaccio ("divina") • "Errors" already in the oldest dated manuscript (Landiano, 1336)
8. Critical tradition • First scientific approach by Barbi (1891) ◦ 400 loci critici (~ SNPs vs. whole sequencing) • Casella (1921) proposed the first tree(s) • Petrocchi (1965) published the most important edition ◦ Excludes everything that might have been influenced by Boccaccio ◦ Stemmatic, but also practical reasons • Sanguineti (2001) proposed a new tree ◦ Collatio of the full text, but only seven manuscripts ◦ Rb and Urb have a new placement • Shaw (2011) contrasts Sanguineti with a digital edition and phylogenetic analyses
9. Casella's tree(s)
10. Critical tradition • First scientific approach by Barbi (1891) ◦ 400 loci critici (~ SNPs vs. whole sequencing) • Casella (1921) proposed the first tree(s) • Petrocchi (1965) published the most important edition ◦ Excludes everything that might have been influenced by Boccaccio ◦ Stemmatic, but also practical reasons • Sanguineti (2001) proposed a new tree ◦ Collatio of the full text, but only seven manuscripts ◦ Rb and Urb have a new placement • Shaw (2011) contrasts Sanguineti with a digital edition and phylogenetic analyses
12. Critical tradition • First scientific approach by Barbi (1891) ◦ 400 loci critici (~ SNPs vs. whole sequencing) • Casella (1921) proposed the first tree(s) • Petrocchi (1965) published the most important edition ◦ Excludes everything that might have been influenced by Boccaccio ◦ Stemmatic, but also practical reasons • Sanguineti (2001) proposed a new tree ◦ Collatio of the full text, but only seven manuscripts ◦ Rb and Urb have a new placement • Shaw (2011) contrasts Sanguineti with a digital edition and phylogenetic analyses
14. Critical tradition • First scientific approach by Barbi (1891) ◦ 400 loci critici (~ SNPs vs. whole sequencing) • Casella (1921) proposed the first tree(s) • Petrocchi (1965) published the most important edition ◦ Excludes everything that might have been influenced by Boccaccio ◦ Stemmatic, but also practical reasons • Sanguineti (2001) proposed a new tree ◦ Collatio of the full text, but only seven manuscripts ◦ Rb and Urb have a new placement • Shaw (2011) contrasts Sanguineti with a digital edition and phylogenetic analyses
16. Many questions • Can we reproduce Petrocchi's tree with Bayesian analysis? ◦ Can we support it by using the entire text instead of the 400 loci? • Can we support Sanguineti's tree, especially the relationship between Urb and Rb? • Can we reproduce Shaw's findings with Bayesian analysis? • Is there evidence for an independent tradition for each cantica? • Can we start building a pipeline for: ◦ Incorporating more manuscripts of the Divine Comedy? ◦ Studying other works? Papyrus 52 (oldest fragmentary manuscript of the New Testament)
17. Data and analysis • Collecting and expanding data from Shaw's edition • 34 taxa ◦ 6 (+1) witnesses ◦ layers of correction independently coded ◦ 3 critical editions (Petrocchi, Leonardi, Sanguineti) • 94,782 characters (34,417 sites, 1,487 patterns) • Most characters with either a single state or two states ◦ Not considering orthography ◦ However, in some cases over 10 states • Initial exploratory analyses with Neighbor-nets; Bayesian with BEAST2 ◦ Only seven main manuscripts / all witnesses ◦ The entire text / each cantica independently ◦ 400 loci / the entire text
18. Example of the raw source data provided by Shaw (2011).
19. NEXUS
21. Covarion, birth-death, strict clock, global variation, MCMC length 25M, all taxa
22. Covarion, birth-death, strict clock, global variation, MCMC length 25M, selected taxa
23. Covarion, birth-death, relaxed clock, rate variation, MCMC length 500k, selected taxa
24. ASR is hard(er) • In the first canto (NEXUS character I_01_048_7), the author describes a terrifying she-wolf walking towards him a. parea che l'aere ne temesse ("as if the air afraid of it [=the she-wolf]") b. parea che l'aere ne tremesse ("as if the air was shaken by it [=the walking]") • Experts' consensus is that it is a "facilitation" of the original, employing the Latinism tremere (Italian has tremare) ◦ Common principle of lectio difficilior potior (the most difficult state is the strongest) ◦ Used by the author elsewhere (e.g. Letter VI, 24, and Rime IV.1-2, where the air is "shaken by a person walking") ◦ Literary critics agree that it fits the tone much better
25. Higher goals • Problem of the Synoptic Gospels • Evolution of religious texts in light of history. For example, in earlier manuscripts: ◦ Mark ends with the death of Jesus (Mark 16:9-20 not found) ◦ Luke 23:34 ("Forgive them, for they know not what they do") is missing ◦ Paul's injunction to women to be "silent" and "subordinate" (1 Cor 14:34-35) is missing
26. References ALAGHERII, Dantis (2001). Comedìa. Edited by Federico Sanguineti. Firenze: Edizioni del Galluzzo. ALIGHIERI, Dante (1994). La Commedia Secondo L’antica Vulgata: Introduzione. Edited by Giorgio Petrocchi. Opere di Dante Alighieri v. 1. Firenze: Le Lettere. BORDALEJO, Barbara (2006). Modern Genetic Methods: the New Stemmatics. http://www.textualscholarship.org/stemmatics/index.html. Accessed on December 4, 2019. CERQUIGLINI, Bernard (1989). Éloge de la variante: histoire critique de la philologie. Aux Travaux. Paris: Éditions du Seuil. HOWE, Christopher J.; CONNOLLY, Ruth; WINDRAM, Heather F. (2012). “Responding to Criticisms of Phylogenetic Methods in Stemmatology.” Studies in English Literature 1500-1900 52 (1): 51 – 67. LEONARDI, Anna M. C. (1991). “Introduzione.” In La Divina Commedia. Milano: Arnoldo Mondadori Editore. MACÉ, Caroline; BARET, Philippe V. (2006). “Why Phylogenetic Methods Work: the Theory of Evolution and Textual Criticism.” Linguistica Computazionale 24: 89 – 108. ROBINSON, Peter (2016). “Four Rules for the Application of Phylogenetics in the Analysis of Textual Traditions.” Digital Scholarship in the Humanities 31 (3): 637 – 651. SHAW, Prue (2011). Commedia: a Digital Edition. Birmingham: Scholarly Digital Editions. TROVATO, Paolo; TONELLO, Elisabetta (2016). “Premessa ai saggi di edizione e di commento di Inferno XXXIV.” Commedia (2). Padova: Libreriauniversitaria.it.
27. Thank you! Main findings so far: • Phylogenetic analysis supports Shaw (and Petrocchi) more than Sanguineti ◦ But it is not so clear-cut • No support for an independent tradition of each cantica • Loci critici seem to perform better than random selections of equivalent size • Evolutionary models are not suitable out-of-the-box ◦ Date-calibration can worsen the results • Probabilities of substitution might depend on character as well as value