Saggi
T HE T E X T UAL T RADI TIONOF DANTE’S
« C OMME DI A» AND THE«BARBI LOCI»
PETERROBINSON
ThetextualproblemwiththeCommediaissimplystated.Asitsmost
recenteditor,PrueShaw,putsit:«Thefirstproblemfacinganyeditor
ofDante’sCommediaistherichnessofthemanuscriptrecord–around
600copiesifwecountonlycompletetextsofthepoem,morethan800if
weincludepartialandfragmentarycopies».1Intermsofsize:noeditor,
noteamofeditorsworkingwiththetraditionalmethodsofItalianphilologycouldeverfindtheresourcestolookateveryreadingineveryone
ofthe14233linesoftheCommediaineveryoneofthe800plusmanuscriptsandsignificantearlyprinteditionsofthepoem.
Noristheproblemjustthenumberofmanuscripts.Itisknownthat
threeearlyandveryinfluentialrescensionsofthewholetextweremade
byGiovanniBoccacciobetween1355and1373,andthatallthreeprimaryBoccacciocopies,thoughbasedonthestill-extantVaticanmanuscriptknowntoCommediaeditorsas«Vat»,introducedmanyreadings
frommanyothermanuscripts:whateditorsknowas«contamination».
Fromthatpointon,themanuscriptsshowsuchamixtureofreadings
thatstandardstemmaticprocesses,dependingontheorderlycopyingof
EarlierversionsofthispaperwerereadatmeetingsoftheStudiaStemmatalogicaworking
group,convenedbetween2010and2012byTuomasHeikkilaoftheUniversityofHelsinki.
Iamgratefultomembersoftheworkinggroupfortheircommentsonhemethodologyof
thepaper,andespeciallytoStevenStevenJ.SchwagerandTeemuRoosfortheirdetailed
helpwithhypergeometricdistribution.IamgratefulalsotoPrueShawforherhelpwith
thediscussioninthelatterpartofthepaperoftheinfluenceofContinionItaliantextual
scholarship–and,asalways,fortheprivilegeofworkingsocloselyandsolongwithher.
1
«Introduction»inPrueShaw,ed.,DanteAlighieri.Commedia.ADigitalEdition,Birmingham-Florence,ScholarlyDigitalEditions-SISMEL-EdizionidelGaluzzo,2010.
2
PeterRobinson
readingsfromonemanuscriptintoanother,appearimpossible.2Evenif
onecouldrecordeveryreadingineverywitnessthereiscommonagreementamongmanyscholarsthattraditionalstemmaticanalysis,asidentifiedwithLachmannandthefollowersofthemethodnamedforhim,
cannotdealwithcasesofcontaminationandhenceitisimpossibleto
constructausefulgenetichypothesisconcerningtherelationsamong
themanymanuscriptsoftheCommedia.3Itappears,then,thatweare
facedwithanimpossiblesituation.Thefundamentalimportanceofthe
CommediatoItalianandworldliteraturemakesitimperativethatDante’smaster-workbeedited,andeditedtothehighestpossiblestandard.
Yet,eitherdifficulty–thesheersizeofthetraditionortheprevalenceof
contamination–wouldonitsownmakeitimpossibletoachievewhat
onecouldarguearethetwominimalrequirementsofascholarlyedition:first,thatitexaminealltheevidenceinallthewitnesses;second,
thatitderivefromallthisevidenceanunderstandingofthewholetraditionwhichcanthenbeusedbytheeditortoidentifywhichmanuscriptsandwhichreadingswithinthemaremostlikelynearesttoDante’sowntext.4
OverthelastcenturiesofDantescholarship,scholarshavetriedseveralroutespastthisimpossibility.GiorgioPetrocchi,editorofthemost
signficantmoderneditionoftheCommedia,addressedbothproblems
withasinglestrategy:thatofbasinghiseditionoftheCommedia«secondol’anticavulgata»onlyonthe27manuscripts(twenty-fourcomplete,threefragmentary,andcountingMartini’scollationoftheAldine
editionasamanuscript)acceptedatthetimeofhiseditionasdating
securelybefore1355.5Thisapproachbothreducedthenumberofmanuscriptswhichheneededtoexaminetoamanageablenumber,andalso
greatlyreducedtheproblemofcontaminatedmanuscriptsbylooking
onlyatmanuscriptswrittenbeforeBoccacciostartedworkonhisrecenIndeed,manuscriptsdatablefrombefore1335alreadyshowclearsignsofcontamination:thatis,theimportofreadingsfrommanuscriptsotherthantheexemplar.The
Landinomanuscript(La),datedto1336,hasreadingsscrapedawayandotherssubstitutedacrossitswholelength.Theevenearliermanuscript,notnowextant,datedto13301331andusedbyLucaMartiniinhiscollationofthe1515Aldinetext,alsoincluded
readingsfrommanuscriptsotherthantheexemplar(Shaw,«Introduction:Overview»).
3
AmongMiddleEnglishtextualscholars,theviewthatnousefulgenetichypotheses
canbecreatedforlargemanuscripttraditionshasbecomesowidelyacceptedastocount
asgospel:seeGeorgeKane’seditionsofPiersPlowmanandhisassaultonJohnManly
andEdithRickert(whodidthinktheycoulddisentanglethe80-plusmanuscriptsofthe
CanterburyTales)inPaulRuggiers,ed.,EditingChaucer:theGreatTradition,Norman,
Oklahoma,PilgrimBooks,1984.
2
ThetextualtraditionofDante’sCommedia
3
sionsoftheCommedia.Petrocchistatedveryclearlythatnoeditionbased
onsuchasmallselectionofmanuscriptscouldclaimtobe«un’edizione
critica»,whichmustbebasedonallthemanuscriptevidence,notjust
someofit.6Accordingly,heclaimedonlythemoremodesttitleof«La
Commediasecondol’anticavulgata»forhisedition.Nonetheless,healso
arguedthateventhoughhistextwasbasedonlyonthissmallfractionof
themanuscripts,allthelatermanuscriptswouldnotcontributeasingle
readingnotalreadypresentinorreadilydeduciblefromtheanticavulgatamanuscriptstoacriticaltext.Therefore,hedeclared,asoundeditorialtextbasedonlyontheanticavulgatamanuscriptswouldbeidentical
tothatbasedonthewholetradition,andsohistextcouldstandforthe
wholetradition,despitetheprovisionalstatusimpliedbythe«secondo
l’anticavulgata»qualification.
Petrocchi’sapproachevades,ratherthansatisifies,theneedtoform
aviewoftheentiremanuscripttradition.Indeed,hisassertionthata
soundeditioncouldbebasedonthese27manuscriptsalonecouldonly
bevalidatedbylookingatalltheevidenceinallthemanuscripts.But
howcanthisbedone,givenovereighthundredmanuscripts,morethan
14000linesoftextand100,000wordsinacompletemanuscript?Before
Petrocchi,theItalianDantescholarMicheleBarbi,thenatthebeginningofaremarkablefifty-yearphilologicalcareer,andattheinstigation
ofthreeseniorDantescholarsassociatedwiththenewly-foundedSociThispessimisticview,thatnogenealogicalrepresentationoftherelationships
amongthemanuscriptsoftheCommediaispossible,wasreachedbythenineteenthcenturyEnglishscholarEdwardMoore,whoexaminedhundredsofmanuscriptsofthe
Commediaonlytoconclude«Myownbeliefwouldbethatowingtothecomplicated
intermixtureoftexts,suchagenealogynevercanbeconstructed.»(EdwardMoore,
ContributionstothetextualcriticismoftheDivinaCommedia,includingthecompletecollationthroughouttheInfernoofallthemss.atOxfordandCambridge,Cambridge,CambridgeUniversityPress,1889,p.xxxi).ShawcitestheGermanscholarWitteasimplicitlyreachingthesameconclusionintheProlegomenatohis1862editionofthepoem,
basedonhisownindependentexaminationandcollationofhundredsofmanuscripts
(«Introduction»,n.22).
5
DanteAlighieri,LaCommediasecondol’anticavulgata,acuradiGiorgioPetrocchi,
inLeoperediDanteAlighieri.EdizioneNazionaleacuradellaSocietàDantescaItaliana,
VII,Milano,1966-1967,4vols.NotethatinthisarticleallreferencestoMart,intermsof
readingsfoundandanalysisbasedonthosereadings,aretotheMartinicollationexcept
whereotherwisespecified.
6
«Aifinidiun’edizionecritica,eperchétaleveramentesia,sidevetuttoratenerfede
allaleggedipartenzaedifondod’ogniricercatestuale:l’interrogazioneintegraledellatradizione.»GiorgioPetrocchi,«Proposteperuntesto-basedellaDivinaCommedia»,Filologiaromanza,II(1955),pp.337-365[p.343],citedbyShaw,«Introduction:Overview».
4
4
PeterRobinson
etàDantescaItaliana,producedalistoflinesintheCommediawhich
hejudgedcriticalfortheestablishmentofmanuscriptrelationsacrossthe
wholetextoftheCommediaandallitsmanuscripts.Theseareknown
asthe«400loci»:infact,thereare396ofthem.7Barbiexplainsinalater
articlethattheselociwerechosenonthebasisofconsiderableexperience
workingwithCommediamanuscriptsintheFlorentinelibraries,andon
considerationofthesignificanceofparticularvariantreadings.8Aswell
asreducingtheamountofeffortrequiredtosurveythewholetradition(oneneedlookatonly396lines,notat14233),thisalsoofferedthe
promiseofacollaborativeapproach:theworkcouldbedividedamong
manyscholars.Accordingly,specialformswereprintedoff,tobedistributedtoscholarsandsoenableacompletesurveyofthesekeylinesin
everymanuscripteverywhere.AsShawrelates,theresponsewasdisappointing:someelevenscholars,includingBarbihimself,examinedafew
manuscriptsandreportedtheirfindings.AmongtheseelevenwereGuiseppeVandelliandMarioCasella:yetwhenthesetwoscholarscameto
producetheirowneditionsoftheCommediain1921and1923respectively,neitherusedthecollationoftheBarbilociasthebaseforananalysisofthewholetradition,andhencetheiredition.
Aswehaveseen,thenextgreateditionafterVandelli,thatofPetrocchi
in1965,alsomadenouseoftheBarbiloci,choosingadifferentrationale
foraselectionofmanuscriptsonwhichtobasetheedition.Lanza’s1995
editiontookanevenmoreextremeapproachtotheproblemofhowto
deriveasingletextfromavastnumberofmanuscripts:heusedjustone
manuscript,theTrivulziano1080(«Triv»),oneoftheveryearliestsurvivingFlorentinemanuscriptsandlongfamousforthequalityofitstextand
thebeautyofitsdecoration.Inthischoice,Lanzawasactingconsciously
inthetraditionof«best-textediting»associatedwithJosephBédier.FollowinghisworkonLeLaidel’ombre,Bédierassertedtheimpossibilityof
creatingausefulgenetichypothesisaboutanylargeandcomplexmanuscripttradition.9Inhisview,Lachmannianstemmaticsisanimpossibil7
ThethreescholarswereAdolfoBartoli,AlessandroD’Ancona,IsidoroDelLungo,
writingin«Perl’edizionecriticadellaDivinaCommedia»,BullettinodellaSocietàDantescaItaliana,5-6(1891),pp.25-27,andfollowedbyBarbi’s«Canonediluoghisceltiper
lospogliodeimss.dellaDivinaCommedia»onpages28-38.AlltheBarbilociarelisted
inAppendixAofShaw’sedition,withconvenienthypertextlinkstotheedition’scollationofeachlineinthesevenmanuscriptsandtwoeditionsincludedwithinit.
8
«Fufruttodilunghistudi,efissataquindinonapriori,cioèacaso...»;Barbi,«Ancora
sultestodellaDivinaCommedia»,Studidanteschi,XVIII(1934),p.56.
9
«LatraditionmanuscriteduLaidel’ombre»,Romania,54(1928),pp.161-296,321356;reprintedParis,E.Champion,1970.
ThetextualtraditionofDante’sCommedia
5
ityinsuchcircumstances.Theonlyrecoursethenistoidentifythebest
singlewitnesstothetext,bywhatevercriteriaareavailable,andusethat
asthebaseforanedition.AsLanzaalsoargued,suchaneditioncan(if
rigorouslyfaithfultothesource)presentatextwhichactuallyexisted
atasinglehistoricmoment,unsulliedbyeditorialimpulseorlinguistic
revision.Infact,Lanzadidpermitemendationsfromothermanuscripts
intohisedition,butonlyunderstringentcircumstancesandfromavery
limitedrangeofotherFlorentinemanuscripts.Ofcourse,thisapproach
freestheeditorcompletelyfromtheneedtosurveytheentiretradition,
whetherbyusingtheBarbilociorbyanyothermeans.
Around1996,then,Barbi’slocilookedlikearoad-not-takeninCommediatextualscholarship:aninterestingideabutneverimplemented.
However,in2001aneweditionoftheCommediaappeared,editedby
FedericoSanguineti,inwhichtheBarbilocimovedbacktothecentre
stageofCommediatextualscholarship.10Thiseditionwasstartlingfor
severalreasons.First,wherePetrocchi,Lanzaandothereditorsagreed
thatitwasimpossibletoformausefulgenetichypothesisoftheentire
tradition,Sanguineticlaimedthatitwaspossibletoachievethisforthe
Commedia.InoppositiontoBédier’sgloomypessimismaboutstemmatics,SanguinetiassertedthatLachmannianstemmaticscouldbepracticed
rigorouslyandusefullyonthewholeCommediatradition.Second,Sanguinetideclaredthatnotonlycouldtraditionalstemmaticsbeappliedto
thewholetradition,buthehaddoneit.HehadlookedatalltheBarbi
lociineveryoneofthe800manuscripts(soachievingonhisown,with
virtuallynosupport,whatscholarshadfailedtoachieveinoverahundredyears),andfromanalysisofthereadingsattheselocihehadcreated
acomprehensiveaccountofthewholetradition,andisolatedjustseven
manuscriptsasnecessaryandsufficientforthecreationofacriticaltext.
Sanguineti’sidentificationofthe«Sanguinetiseven»wasthestarting
pointofShaw’sedition.Itseemedpossibletoproduceadigitaledition
whichwoulddoforthesesevenmanuscriptswhatShaw’seditionofDante’sMonarchiahaddoneforthemanuscriptsofthatwork:thatis,offer
fullmachine-readabletranscriptsofthem,createanextremelyprecise
collationofthetranscripts,andthenusevariousmethods(includingphylogeneticanalysis,derivedfromevolutionarybiology)tocreateaviewof
therelationshipsamongthemanuscripts.11Atfirst,Shaw(andI)planned
10
DantisAlagheriiComedìa,edizionecriticapercuradiFedericoSanguineti,Firenze,
EdizionidelGalluzzo,2001.
11
PrueShaw,ed.,DanteAlighieri,Monarchia.ADigitalEdition,Birmingham-Florence,
ScholarlyDigitalEditions-SocietàDantescaItaliana,2006.SeethereviewarticlebyPaolo
6
PeterRobinson
toworkwithSanguinetiinthemakingofthisdigitaledition,andthere
wereseveralsubstantivediscussionstowardsthisaim.However,Sanguinetileftthepartnershipatanearlystage,andalmostalltheworkofthe
editionproceededwithouthisinvolvement.Inturn,theaimoftheedition
changed:itbecameatestofSanguineti’sargumentsabouttherelation-
shipsamongthesesevenmanuscripts.ShawexaminesSanguineti’sclaims
inconsiderabledetail,andthereisnoneedheretodomorethansummarizehercriticismsofthem,bothintermsofhisconclusionsandof
hismethodology.Briefly:itappearsfromheranalysis,basednotonany
selectionofvariantsbutontheentirebodyofvariationintheseven
manuscripts,firstlythatSanguineti’skeyconclusioniswrong,andsecondlythatthemethodologyheusedtoreachthisconclusionwasflawed.
Sanguineti’skeyconclusionisthatonemanuscript,andonemanuscript
alone–VaticanLibraryms.Urbinatelatino366(Urb)–representsapure
lineofdescentfromDante’soriginaltextindependentofeveryother
manuscript.Thushisstemmaofthesevenmanuscripts(figure1):
figure1
Sanguineti’sstemmaofthe«Sanguinetiseven»manuscriptsoftheCommedia.
NotethepositioningofRbassharinganancestorαwiththefivemanuscripts
LauSCMartTrivAshHam,leaving«U»(Urb)asindependentofRbandall
othermanuscripts.
ω
α
[y]
[β]
x
z
a
b
LauSCMartTrivAshHamRbUrb
(L)(M)(T)(A)(H)(R)(U)
Trovato(withanappendixontheShawCommediaedition)in«LaDoppia“Monarchia”di
PrueShaw(conunapostillasullaCommedia)»,Ecdotica,7(2010),pp.193-207.
ThetextualtraditionofDante’sCommedia
7
Here,Dante’soriginalisω,andSanguinetiarguesthatsixoftheseven
descendfromasinglecopyof ω,labelledα,whileUrbisonitsown,des-
cendedindependentlyfromω.Accordingly,theonemanuscriptUrb
has the same authority as the other six combined.Where most editionshavecentredtheirtextaroundtheFlorentinemanuscriptsinthe α
branchofthisstemma(thus,LanzausedTrivashisbase,CasellaLauSC;
PetrocchioftenfavouredreadingsinTrivandMartaheadofthosein
Urb)SanguinetibasedhisentireeditiononUrb.Sanguineti’sargument
dependsonUrbsharingnoancestorbelowωwithanyothermanuscript,andsocontradictsPetrocchi,whoarguedthatRbandUrbshare
anancestor(hise)belowtheoriginal(figure2).AsShawasserts,«the
positionofms.RbinthestemmaiscriticalforSanguineti’sargument:
figure2
PartofPetrocchi’sstemmaoftheanticavulgatamanuscriptsoftheCommedia.
NotethepositioningofRb,assharinganancestorwithUrb(andMad).
β
e
Rb
Urb
Mad
ifheiswrongaboutthat,tuttocrolla–thewholeedificecollapses.»Her
analysis,supportedbyphylogeneticanalysis(seefurtherbelow),isdecisive:Petrocchiisright,RbandUrbdoshareanancestor,andSanguineti
iswrong.Further,onecantraceSanguineti’serrordirectlytohismeth-
8
PeterRobinson
odology.AlthoughSanguineticlaimedthathebasedhisanalysisonthe
396Barbiloci,Shawpointsoutthatheactuallyusesveryfewofthemto
supporthisconclusions.Sanguinetiidentifiesoverfourhundredmanuscriptsasmembersofasinglegroup(«lacosiddetta“tradizioneβ”»)on
thebasisofjustfourreadings.Further,oneofthesereadings(Par.xxiii
103,spiroforgiro)hasthecharacterofaclassicpolygeneticerror,which
couldreadilyoccurindependentlyinunconnectedscribalcopies.Elsewhere,hejustifiesdiscardingtheevidenceofthislargegroup(expanded
toover600manuscriptsonthebasisofaveryfewreadings)becauseitis
contaminatedbyreadingsdrawnfromthedistincttraditionswhichhe
seesunderliningthesevenmanuscripts:thus,theappearanceofasingle
readingfromRbinsomeofthese600manuscripts(aleforaeratPurg.ii
35)isusedtoarguethatthisgroupiscontaminatedfromRboramanuscriptcloselyrelatedtoit.
Manifestly,argumentsbasedonjustafewreadingsinatextofover
100,000wordswilllackconviction:areasonwhyBarbisuggestedanalysisofsome400linesoftext,notjusttwentyorso.Ideally,onewould
baseanalysisonthewholetext,oneveryvariantateverywordinevery
line.Todothisforall800manuscriptswouldbeimpossiblewiththe
resources available to any current project; but one could do it for a
smallernumber:indeed,forthesevenmanuscriptsidentifiedbySanguineti,andthiswaswhatShawdid.Thebaseofhereditionwasacompletetranscriptionofthesevenmanuscripts,andcreationofaveryprecisecollationofeverywordofeveryoneofthesesevenwitheachother,
and with two major edited texts: those of Petrocchi and Sanguineti.
Whiletheedition’seventualaimwastousethiscollationandtheanalysisbuiltonittoexploreSanguineti’shypothesis,asexplainedinthe
lastparagraphs,thissamecollationcanbeusedforanotherpurpose:to
testtheefficacyoftheBarbilocifortheirdeclaredaim(asacceptedby
Sanguineti),asabaseforananalysisoftheentiretradition.Ofcourse,
inthiseditionwehaveonlysevenmanuscripts.However,thekeyplacein
thetraditionofatleastsixofthese,acceptedbyeveryscholarinthe
lastcentury,makesthesemanuscriptsagoodplacetostart.Moreover,
aswehaveseen,Sanguinetibasedhisidentificationofthesesevenmanuscriptsascriticaltotheunderstandingofthewholetraditionofthe
Commediaonhiscollationofthese396lines.Accordingly,oneshould
expectthattheselocishouldbepre-eminentlyinformativeintheanalysisofthesesevenmanuscripts.
Shaw’seditionproceededasfollows.First,anexactword-by-wordcollationofthewholetextinthesevenmanuscriptswascreatedasfollows:
ThetextualtraditionofDante’sCommedia
9
1.Allsevenmanuscriptsweretranscribedinfull,wordbywordand
linebyline,intomachine-readableform.
2.Thetranscriptionrecordedlayersofscribalrevisionineachmanuscript,sothatonecoulddistinguishatanypointwhatwaswrittenby
theoriginalscribe,changedbyhimorher,thenchangedagainbylater
scribes.ThisisparticularlyusefulwithrespecttoMart,whereitisactuallythereadingsrecordedbyLucaMartinifromanow-lostmanuscript
asalterationstotheAldineeditionwhichareofprimaryinterest.Thus,
inPurg.vii.51,wheretheAldineeditionprints«opursarria»,Martini
writes«ononsaria»intherightmargin(figure3),bothreadingsare
recordedinourtranscription.Further,thelayersofreadingsaremarked
sotheycanbecomparedseparately.
figure3
Martini’sannotationstotheAldineeditionatPurg.vii.51.
figure4
The«Literal»viewofthetranscription,showingboththeoriginaltextandMartini’sannotationsasmarkedintherightmargin.
figure5
The«AldineOriginal»viewofthetranscript,showingthetextasoriginally
printed.
figure6
The«Martini’sCollation»view,showingthetextwithMartini’sannotations
applied.
10
PeterRobinson
3.Thetranscriptswerethencomparedwithoneanother,andwith
thetextsofthePetrocchiandSanguineti,byShawandherteam,usingthe
computerprogramCollate.Thecollationremovedorthographicand
spellingvariants,ascanbeseeninthiscollationofthesecondwordof
thefirstlineoftheInfernocanto1showninfigure3:
figure7
Collationofmezzo,Inf.i.1.
Here,theeightdifferentspellingsofmezzointheninewitnesses(seven
manuscriptsplusSanguineti[FS]andPetrocchi[PET])areregularized
tomezzo,sothatnovariantshowsinthisline.
Thecollationalsopreservedinformationaboutthelayersofscribal
revision,sothatonecouldcomparethedifferentstatesofthetextwithin
amanuscript,andbetweenthatmanuscriptandothers.Thus,thecollationatPurg.vii.51infigure8distinguishestheAldinetext(«Martorig»)andMartini’scollation(«Mart-c2»),andshowstoothatthereading«opur»isfoundinUrb,andhenceinSanguineti(«FS»),whilefour
othermanuscripts(AshHamRbTriv)havethesamereadingasMartini
(thisisalsothereadingofPetrocchi)andLauSCalonereads«ouer.»
figure8
Thecollationof«onon»inPurg.vii.51.
Attheendofthisprocess,acompleterecordofeverysignificantvariantateverywordoftheCommediaateverylayerofwritinginthese
sevenmanuscriptswasavailable.
Thenextstepwastoanalyzethisrecordofvariationtoseewhatcould
bededucedaboutthemanuscriptrelations.Ouranalysisfollowedtwo
paths.First,Shawusedthetraditionalmeansofphilology,scrutinizingall
thevariantsbearingonherargument,categorizingthem,andseeking
explanationswhichtakeaccountofthefullrangeofvariants.Thus,in
thekeysection«ThepositionofRb»shelooksattwenty-eightvariants
ThetextualtraditionofDante’sCommedia
11
discussedbyPetrocchi,andthensupplementsherdiscussionoftheseby
examinationofafurther76variants.Herargumenthingesnotonlyon
thenatureofthevariantsinthemselves,butalsointheconsistentagreementofUrbandRbagainstall,ornearlyall,theothermanuscripts:
Someofthesecategoriesarenotespeciallysignificantinthemselves(manyof
themareincludedinBrandoli’scategoriesofpolygeneticerror).Singly,they
meannexttonothing.Butitisthepresenceofalongseriesofthemuniformly
rightacrossthetextinaverysmallnumberofmanuscriptswhichisstriking
(andthisissurelywhatPetrocchi’sphrase«foltezzadistatistica»referstoatleast
inpart).Itissimplyimpossibletoimaginethatcopyistsworkingindependently
wouldmakepreciselythesesmallchangesatpreciselythesesamepointsright
acrossatextofthislength.Themosteconomicalhypothesisisthattheyareworkingfromacommonexemplarandinheritthesereadingsfromthatexemplar.
figure9
Theunrootedphylogramforthesevenmanuscripts,fromtheShawedition.The
arrowpointingtothenodelinkingallofAsh/Ham,Urb/Triv,LauSCandMart/
Trivhasbeenaddedtothisfigure.
Rb
Allmss,wholetext
LauSC
Urb
Mart
Triv
Ash
————1000changes
Ham
12
PeterRobinson
Inthesecondpathofanalysis,ItookthecompleterecordofvariationcreatedbythecollationandsubmittedthattoanalysisbytheprogramPAUP(«PhylogeneticAnalysisUsingParsimony»),usedbyevolutionarybiologiststocreatehypothesesaboutthegeneticrelationships
oforganismsbasedonthecharacteristicstheyshareanddonotshare.
Afullaccountofhowphylogeneticanalysisworks,andwhyandhowitis
suitedtotheanalysisoftextualtraditions,withevidenceofitssuccessful
useinmanycontexts(includingonartificialtraditionsspeciallydevised
totestitsadequacy)maybefoundelsewhere.12figure9givesthe«phylogram»forthewholeCommediageneratedbyPAUPbasedonsome
94,000placesofvariation(typicallyasingleword;butalsophrases)in
thesevenmanuscripts.
Thisrepresentationofmanuscriptrelationsdiffersfromatraditional
stemma(asforexample,thatgivenbyPetrocchi).Firstly,itis«unrooted»:
thatis,itrepresentsthegroupingsofthemanuscriptswithoutanypresumptionastooriginality,orthedirectionofvariation.Secondly,the
manuscriptsareshownasrelatedtooneanotherthroughsharednodes,
withthreepairseachsharingacommonnode(Mart/Triv;Urb/Rb;Ash/
Ham,andtheseventhmanuscript(LauSC)descendingfromanode
onthelinebetweenMart/Trivandthearrowednodelinkingtheother
fourmanuscripts.Therelativelengthsofthelinesbetweennodes,and
betweennodesandmanuscripts,aresignificant,andonecanmeasure
approximatelythedifferencesbetweenthemanuscriptsusingthescale
«1000changes»(or,tenchangespercanto,oneevery14.23lines)onthe
bottomleft.Thus:thelinebetweenthenodejoiningAsh/Hamupto
thearrowednodewhichlinkstotheotherfivemanuscriptsisaboutthe
samelengthasthescaleline(thus,c.1000changes),andapproximately
doublethelengthofthelinefromthearrowednodetothesharedancestorofUrb/Rb(thus,c.500changes),andapproximatelythesameasfrom
thearrowednodetotheancestorofMart/Triv(thus,c.1000changes
again).
Experienceusingthesephylogramshastaughtustobecareful.This
phylogeneticanalysisappearstosuggestthatSanguinetiiswrong:Urb
andRbdoindeedshareacommonancestor,asbothPetrocchiandShaw
assert.However,thesoftware,lefttoitself,willalwaysplaceamanuscriptsomewhereonthephylogram:sometimesthiscansuggestarelaSee,forexample,thepublicatonslistedathttp://www.textualscholarship.org/newstemmatics/bibliography/index.html;notablythecollectionofarticlesinPietervan
Reenen,AugustdenHollanderandMargotvanMulken,eds.,StudiesinStemmatology
II,Amsterdam,JohnBenjamins,2004.
12
ThetextualtraditionofDante’sCommedia
13
tionshipwhichinfactdoesnotexist.Also,thesephylogramscannot
representcontamination.Therefore,wehavetoturntoothermethods
toconfirm,deny,orqualifytherelationshipssuggestedbythephylogeneticanalysis.
InouranalysisofthetraditionoftheCommedia,oneofthe«other
methods»ShawandIusedwasdatabaseanalysisofthedistribution
ofvariantsacrossthemanuscriptsandacrossthewholelengthofthe
tradition.13Ourhypothesisisthis:considerthemanuscriptsdescendingdirectlyfromasinglenode(forexample,UrbandRb).Ifthemanuscriptsdescendeddirectlyfromthissinglenodereallyshareacommon
ancestorbelowthearchetype,thenoneshouldbeabletoidentifyaset
ofvariantslikelytohavebeenintroducedintothatsharedancestor,and
thendescendingtothosemanuscripts.Thus,thereshouldbevariants
whichsatisfythefollowingfourconditions:
1.Theyshouldbepresentinthemanuscriptsdescendeddirectlyfrom
thesharedancestor(thus,inbothUrbandRb);
2.Theyshouldbelikelynottohavebeenpresentinthearchetype;
3.Theyshouldbefoundrarelyinothermanuscriptsoutsidethose
descendedfromthesharedancestor(thus,rarelyinanyofAshHam
LauSCMartTriv);
4.Thereshouldbeasignficantnumberofsuchvariants.
Weusedadatabase-likesearchtool,«VBase»,tofindthevariantswhich
satisfiedtheseconditions.ThistoolisavailableontheDVDpublication,
whereitispre-suppliedwithkeysearcheswefoundusefulinthecourseof
ouranalysis.Hereisoursearchforvariantslikelytohavebeenintroduced
bythejointancestorofUrb/Rb,andwhoseexistencethereforesupports
Petrocchi’sviewthatUrbandRbshareanancestorbelowthearchetype,
andcontradictsSanguineti’sviewthattheydonot:
Thedatabaseanalysiswascarriedoutbymyself,employingallthedatagenerated
bythetranscriptionandcollationdirectedbyShawandusingthecriteriabasedonthe
Shawanalysis.ShawandIdiscussedandreviewedtheresults,andjointlywroteanarticle«PhylogeneticAnalysis»,includedintheShawedition,drawingheavilyonthedatabaseanalysis.Henceforth,thisdatabaseanalysisisdescribedasthe«Shaw/Robinson»
analysis,andthevariantlistsproducedbyitofthedistinctgroupsasthe«Shaw/Robinson»lists.
13
14
PeterRobinson
figure10
VBasesearchforvariantsevidencingacommonancestorbelowthearchetype
forUrbRb.
ThefirstlineofthissearchrequestsvariantsonlypresentinbothRb
andUrb,andhencelikelytohavebeenpresentintheircommonancestor(condition1,above).Thenextlineasksonlytoseethosevariants
presentinfewerthanfive(thatis,inUrbRbandnomorethantwo
other)witnesses,andhencenotlikelytohavebeenpresentintheancestorofthewholetradition(condition2).ThelikelihoodisthatifavariantispresentinthreeormoreotherwitnessesbesidesUrbandRb,then
thatvariantwouldveryprobablyhavebeenpresentinthearchetype.
Indeed,ifoneincreasesthenumberfrom<5to<6,ormore,weseethe
searchreturningmoreandmorevariants,asitcatchesreadingslikely
presentinthearchetype(thus:488readings<6;806readings<7;1881
readings<8;6109<9).Accordingly,thisrestriction(returningonlyread-
ThetextualtraditionofDante’sCommedia
15
ingsinfewerthanfivewitnesses)satisfiesthesecondconditiongiven
above.NotethatbecauseweincludetheSanguinetiedition(«FS»)inour
collation,andSanguinetifrequentlychoosestheUrbreadingoverthat
ofothermanuscripts,asinallthreecasesfromInferno1givenhere,the
restrictiontofiveiseffectivelyUrb+RB+FSplusoneothermanuscript
only(infact,fully247ofthese308readingsreturnedbythissearchare
presentinFS).
Thenextcondition(condition3)isthatthereadingsevidencingan
exclusivecommonancestorforUrbandRbshouldberarelyfoundin
anyothermanuscript.Thistablegivesthenumberofoccurrencesofany
ofthese308readingsacrossallthewitnessesherecollated:
table1
OccurrenceofUrb/Rbreadingsinallwitnesses.
Manuscript
NumberofUrb/Rb
readings(max.308)
PercentofUrb/Rb
readingspresent
Ash
40
13
Ham
46
15
LauSC
11
4
Mart
4
1
Rb
308
100
Triv
9
3
Urb
308
100
FS
247
80
PET
19
6
Becauseoftheoperationofco-incident(orconvergent)variation,it
istobeexpectedthatbysimplechance,aproportionofreadingsintroducedbyonescribeinonecopyingmightbefoundinaquiteunrelated
copymadebyadifferentscribe.Onemighthypothesizethatinanytwo
actsofcopying,thedifferentscribesmightintroducethesamemistake
sometenpercentofthetime,orinoneintenvariants.Onthiscalculation,theincidenceofthe308Urb/RbinallexceptFSAshHamcanbe
explainedassimplecoincidence–andthehighnumberofUrb/RbvariantspresentinFSresultsfromSanguineti’sdecisiontobasehisedition
16
PeterRobinson
onUrb.ThenumberofUrb/RbvariantsinAshandHamismarginally
higherthanonemightexpect,andthismaybetheresultofcontaminationbetweentheUrb/RbandAsh/Hamgroups(seebelow).Accordingly,thisgroupofvariantssatisfiesthethirdconditiongivenabove:
readingsfromthisgrouparerarelyfoundinothermanuscripts.
Thelastconditionisthatthenumberofvariantsshouldbe«significant»:thatis,notjustarandomagreementbetweenthesetwowitnesses
butthelikelyresultofshareddescentfromasingleancestorcontaining
thesereadings.Again,onecanuseVBasesearchestoexplorewhatlevelsof
agreementbetweenanytwowitnessesmightbetheresultofco-incident
variation.Fromthephylogramgiveninfigure9,onemayhypothesize
thatthefollowingpairsofmanuscriptsdonotshareacommonancestorbelowthearchetype:Rb/Mart,Rb/Triv,Urb/Mart,Urb/Triv,Urb/
LauSC,Rb/LauSC,Rb/Ham,Urb/Ham,Rb/Ash,Urb/Ash,Ash/LauSC,
Ham/LauSC,Mart/Ash,Mart/Ham,Triv/Ash,Triv/Ham,Mart/LauSC,
Triv/LauSc.HerearetheresultsofVBasesearchesonthesepairsofmanuscripts,usingthesamecriteriaemployedforthesearchgiveninfigure10:
table2
Numberofreadingsfoundinhypotheticallyunrelatedwitnesspairs,likelytobe
theresultofchanceagreement.Allsearcheswerefor:witnesspair+<4ofallwitnesses,i.e.witnesspair+oneother,exceptforpairsincludingUrb(marked*),
wheresearcheswereforwitnesspair+<5ofallwitnesses,i.e.witnesspair+FS+
oneother.
WitnessPair
NumberofreadingsfoundbyVbase
Rb/Mart
37
Rb/Triv
91
Urb/Mart*
28
Urb/Triv*
58
Urb/LauSC*
99
Rb/LauSC
70
Rb/Ham
219
Urb/Ham*
203
Rb/Ash
201
ThetextualtraditionofDante’sCommedia
Urb/Ash*
167
Ash/LauSC
103
Ham/LauSC
91
Mart/Ash
45
Mart/Ham
57
Triv/Ash
102
Triv/Ham
123
Mart/LauSC
35
Triv/LauSc
107
17
ForallexceptthepairingsinvolvingUrb/Rb/Ash/Ham,thesesearches
typicallyreturnfewerthanonehundredvariants(rangingfrom35for
Mart/LauSCupto123forTriv/Ham).Onemayinferfromthisthatforthis
text,anytwowitnessesmayhappentoshareuptoaround100readings
(oroneacanto)anditwillmeannothingwhatever:theagreementisjust
chance.Conversely,onemaydeducethatchancecannotexplaintwowit-
nessessharing308variants(orthreeacanto):thisislikelytobethe
resultofshareddescentfromacommonancestor.Thus,these308readings may satisfy the fourth condition: their number is significant of
morethanchanceagreement.
IexceptherethepairingsinvolvingUrb/RbwithAsh/Ham.Thenumber
ofagreementsbetweenanytwooftheserangesfrom167(Urb/Ash)to
219(Rb/Ham).Thisseemshigherthanonemightexpectfromrandom
agreement,andwerecallthatthenumberofUrb/Rbreadingsfoundin
AshandHamgiveninTable1wasalsoratherhigherthanonemight
expectfromrandomagreement.ShawandIexaminethisquestionin
our«Phyologeneticanalysis»argumentandsuggestthatthisagreement
acrossthepairsAsh/HamandUrb/Rbislikelytobetheresultofcon-
tamination, particularly in Inferno 1-7 and the last thirteen canti of
Paradiso.
Tosummarize:bythisprocessofVBasesearching,builtuponthe
resultsofthephylogeneticanalysis,weisolatedalistof308variants
whichwebelieveareevidenceofafamilyrelationshipbetweenthetwo
manuscriptsRbandUrb.Thephylogram(figure9)suggeststhatthere
aretwoothermanuscriptpairsamongtheseven:Ash/HamandMart/
Triv.Further,itmaybethatthefivewitnessesAsh/Ham/Mart/Triv/
18
PeterRobinson
LauSC(thatis,thetwopairsAsh/HamandMart/TrivplusLauSC)may
alsodescendfromasinglecopybelowthearchetype,andthereforealso
formafamilygroupingwithinthetradition.FollowingPetrocchi,we
callthisgroupα.Forallthesegroupings,wecarriedoutsimilarVBase
searchestothatwedescribedaboveforUrb/Rb,identifyingforeach
asetofvariantscharacteristicofeachgroupandlikelytohavebeenintro-
ducedbytheexclusivecommonancestorofeach.Thegroups(hereafter,theShaw/Robinsonlists),thenumberofvariantswefoundforeach
group,andthenumberoflinesinwhichthesevariantsoccur(rather
fewerthanthenumberofvariants,assomelineshavemorethanone
variant)aregivenintable3.
table3
Numbersofvariantsevidencinghypotheticalmanuscriptgroups,andnumbers
oflinesinwhichtheyoccur.
Group
No.Variants
No.Lines
Urb/Rb
308
297
Mart/Triv
874
842
Ash/Ham
770
722
α
327
314
total
1885
Takentogether,thisgivesatotalnumberof1885linescontainingvariantswhichouranalysisindicatesaresignificantforestablishingfamily
relationshipsamongthesesevenwitnesses.14Thatis:ofthe14233lines
inthewholeCommedia,itappearsthat1885giveevidenceoffamilyrelationships.EarlierinthisessayIdescribedtheBarbiloci:hislistof396lines
whichhechoseasindicativefortheestablishmentofmanuscriptrelations
acrossthewhole14223linesoftheCommedia.Wehave1885lineswhich
wefindareindicative;Barbisuggests396whichhebelievedtobeindicative.
ThisgivesusastandpointfromwhichwecanassessBarbi’schoice
oflines.BothBarbiandourselves(Shawandmyself)wereseekingthe
samething.Accordingly,onewouldexpectahighdegreeofcorrespond14
Ifoneaddsthenumbersoflines(297+842+722+314)thesumis2175,not1885.
Thisisbecause290linescontainvariantsfrommorethanonegroup.
ThetextualtraditionofDante’sCommedia
19
encebetweenthetwolistsoflines:say,withverymanyofBarbi’s396
linesappearinginourlargerlistof1885lines.Ofcourse,thesamelines
mightappearonbothlistsandmeannothingmorethanchanceagreement.Infact,therearewell-knownprobabilisticmethodsforestimatinginacasesuchasthiswhatproportionofagreementbetweenthelists
mightbesimplechance,andwhatmorethanchance.
Considerthatwehaveanurncontaining14233balls,eachwitha
numberfrom1to14233drawnonit:our14233linesoftheCommedia.15
Wechoose,purelyatrandom,1885ballsfromtheurn:thisistheShaw/
Robinsonlist.Weputall1885ballsback,andwechoose,purelyatrandom,
396ballsfromtheurn:thisistheBarbilist.Now,foranyoneballthat
wehappentochoose,thechancethatitisontheShaw/Robinsonlistis:
1885/14333(=0.13244approx)
Soifwechoose396ballsfromtheurn,thenbysimplechancethenumber
ofballswearelikelytogetwhichareontheShaw/Robinsonlistwillbe:
396x1885/14233(=396x0.13244)=52.446
Thatis:bysimplechanceagreement,ifwechose396linesatrandom
severaltimesover(takingcaretoputbackalltheballsbackintheurn
betweenchoices)wewouldexpecttheaveragenumberoflineschosen
toapproximateto52.446.Saywedidthisexerciseofchoosingtwosetsof
balls(one,1885balls,another396balls)ahundredtimes,twohundred
times,eachtimenotingdownthenumberofballsincommonbetween
thetwochoices.Sometimestherewouldbe45incommon,sometimes
59,sometimes49,sometimes55,etc.:butthemoretimeswechosethe
twosetsofballs,themoretheaveragenumberoflinesinbothlistswould
approach52.446.Hence,52.466isthe«meannumber».
Probabilitycandoevenmoreforus.The«meannumber»isuseful,
aswecanpresumethatifwefindthatthenumberoflinesincommon
onthetwolistsisclosetothisnumber,itismererandomchance.Thus,
Forthediscussionthatfollows,withtheanalogyoftheurncontaining14233balls,
theexplanationofhypergeometricdistribution,andmanyofthecalculationsofprobabilitiesheregiven,IamdeeplyindebtedtoStevenJ.Schwager,nowProfessorEmeritus
intheDepartmentsofBiologicalStatisticsandComputationalBiology,andofStatisticalScience,atCornellUniversity.Ihavetakenovermuchofhisphrasing,andhiskey
statisticalresults.IamgratefulalsotoTeemuRoosoftheUniversityofHelsinkiforhis
correctionsandsuggestions.
15
20
PeterRobinson
ifwefind52linesincommonbetweenthetwolists,thisisjustchance.
Butwhatifwefind51lines,or53?Instinctively,wewillthinkthisis
likelytobechancetoo.Butwhatof70lines,or100,or30?Atwhatpoint
doesitbecomehighlyunlikelythatagivennumberoflineswilloccur
onbothlistsbychance?Probabilitycangiveusanswerstothesequestionstoo.Thissituation,wherewecanassignstatisticalprobabilitiesto
therangeofnumbersofpossiblelinesincommonbetweenthetwolists
(between0,fornonumbersincommon,and396,witheverynumber
ontheBarbilistappearingonourlist)isaninstanceofwhatstatisticianscalla«hypergeometricdistribution»,anditispossibletocalculate
variousprobabilitiesforeachnumberoflinesbetween0and396,thus:
1.Theprobabilitythatexactlythisnumberoflinesshouldbeincom-
moninthetwolists.For52lines,thatis0.0599,orslightlybetterthan
onechanceintwenty(for51itis0.0592,for530.0594)
2.Theprobabilitythatthereshouldbethisnumberoflinesorfewer
incommoninthetwolists.For52lines,thatis0.5102,orjustbetterthan
onechanceintwo–theequivalentoftossingacoinandgettingheads.
3.Theprobabilitythatthereshouldbethisnumberoflinesormorein
commoninthetwolists.For52lines,thatis0.5498,orjustbetterthanone
chanceintwo–again,theequivalentoftossingacoinandgettingheads.
Wecanusethesefigurestotellus,foranygivennumberoflines,not
justwhatthechancesareofexactlythatnumberoflinesoccurringinboth
lists,butalsowhatarethechancesofalesserorgreaternumberoflines
occurring.Thatis:wecanusethehypergeometricdistributiontabletotell
us(say)whatarethechancesoftherebeing30orlesslinesincommonin
thetwolists(thenumberis0.0002,oraround2chancesin10,000),orof70
linesormoreinthetwolists(thenumberis0.0066,orsixchancesin1000).
Evenmoreusefully:wecanexaminethehypergeometricdistribution
toassign«upper»and«lower»bounds.Instatisticalterms,anything
withalessthan0.05probability,thatislessthanoneintwenty,maybe
regardedasunlikely.The«upper»boundwillbethelowestvaluewitha
probabilityoflessthan0.025thattherewillbesomanylinesincommon
betweenthetwolists;the«lower»boundwillbethehighestvaluewitha
probabilityoflessthan0.025thattherewillbesomanylinesincommon
betweenthetwolists.FortheBarbilistof396lines,theprobabilityof
therebeing66ormorelinesincommonwiththeShaw/Robinsonlist
of1885linesis0.0275(for67linesormore,itis0.020):thisgivesan
«upperbound»of66.Theprobabilityoftherebeing39orlesslinesin
commonwiththeShaw/Robinsonlistis0.023(for40linesorless,itis
0.039):thisgivesa«lowerbound»of39.Henceonecandeclarethatthe
ThetextualtraditionofDante’sCommedia
21
chancesarearound19in20thatbysimplechance,therewillbebetween
39and66linesincommononthetwolists.Correspondingly,ifwefind
thenumberoflinesincommonisoutsidetherange39to66,thenitis
unlikelythatthisissimplechance;andthefurtheroutsidethisrange
isthenumberoflinesincommon,themoreunlikelyitisthatthisis
simplechance.
Armed with these formulae, we may look at the correspondences
betweentheBarbilociandtheShaw/Robinsonlistsoflineswefound
significantforthestemmaticanalysisoftheCommedia(Table4).
table4
Numbers of lines on both the Shaw/Robinson and Barbi lists (second last
column),comparedtothenumberexpectedbysimplechance(thirdcolumn),
theupperandlowernumbersconsistentwithchance(<0.05,oroneintwenty),
andtheprobabilityofthatnumberoflinesorgreaterincommon(lastcolumn).
Group
Shaw
Barbiand
Shaw/Rob.
Mean
Lower
bound
Upper
bound
Barbiand
Shaw/Rob.
actual16
Probability
(>=)
Urb/Rb
297
8
3
14
15(6)
0.0194(0.838)
Mart/Triv
842
23
14
34
64(64)
<1x10-12
(superscript-12)
Ash/Ham
722
20
11
30
49(37)
<1x10-10
(superscript-10)
(0.00025)
α
314
9
3
16
19(14)
0.0013
(0.0566)
totals
1885
52
39
66
121
<1x10-13
(superscript-13)
16
ManyoftheBarbilinescorrespondwithmorethanonegroupintheShaw/Robinsonanalysis.Thatis:whileatotalof121linesfromtheBarbilistsappearamongthe
1885linesidentifiedintheShawanalysis,26ofthe121linesappearinmorethanone
oftheShaw/Robinsonlists.Accordingly,oneshouldadjustthecountofcorrespondinglinesforeachgrouptotakeaccountofthisoverlap:theadjustedlinecountsare
giveninparenthesesinthiscolumn,andthecorrespondingprobabilityinparenthesesinthelastcolumn.Toexplain:presuming,asIargue,thatBarbi’schoiceoflineswas
conditionedbyhisawarenessofTriv,thenlineswhichevidencebothMart/Trivand
anothergroupingshouldnotbecountedamonglinesfoundbyBarbiforthatother
grouping.Therefore,oneshouldremovefromtheUrb/Rbcountof15thefourlines
alsofoundintheMart/Trivgrouping,hence11.Trivisalsoaprimewitnessforvariants
foundinthehypotheticalαgroup,andsooneshouldalsoremovethefivelinesshared
byRb/Urbwithα:hence,sixvariants(15-4-5),asgiveninparentheses.Similarly,one
22
PeterRobinson
Lookfirstatthebottom«Totals»lineintable4.Thissuggeststhat
giventheShaw/Robinsonlistof1885linesandtheBarbilistof396lines,
bothdrawnfromthe14233linesoftheCommedia,onewouldexpect
thatbysimplechancealonebetween39and66linesmightappearon
bothlists.ThelastcolumntellsusthatBarbiinfacthas121oftheShaw/
Robinsonlist,considerablyhigherthanonemightexpectfromsimple
chance.Indeed,thechanceoftherebeing121(ormore)linesincommon
is0totwelvedecimalplaces;or,toputitanotherway,roughlyequalto
thechanceofthrowingacoinfortytimesandgettingheadseachtime.
Itcannotbechancethattherearesomanylinesappearingonbothlists.It
mustbebecausebothlistsseektorepresentthesamephenomena,inthis
case:linesusefulforgeneticreconstructionofthetradition.Forthese
121lines,then,Barbi’shypothesisthattheselineswouldbeusefulfor
geneticreconstructionappearsjustifiedbyouranalysis.Accordingly,an
analysisbasedonthe396linesselectedbyBarbiwillgiveabetterpicture
ofthemanuscriptrelationsthan(say)justchoosingany396lines.To
thatextent,onemightindeedusetheBarbilocitowardsageneticreconstructionofthehistoryofthetradition,asBarbiplannedandasSanguinetiattempted.
However,notallgeneticreconstructionsareequal.Acloserlookat
Table4showssomeproblems.Considerthenumbersofvariantsfound
intheBarbilistwhichalsoappearontheShaw/Robinsonlistsforeach
ofthefourgroups.ThenumberoflinesfortheMart/Trivgroupfoundon
theBarbilistismuchhigherthanonewouldexpectbysimplechance:
64,overdoublethenumberwhichcouldbereasonablyexplainedby
chancealone(takingtheupperboundof34asthemostchancemight
reasonablygive).ForAsh/Ham,thenumberissignificantlyhigherthan
chancemightgive(thoughlesssothanfortheMart/Triv).Butforα
thenumberisonlymarginallyhigherthanchancemightsuggest(19,
comparedtoanupperlimitof16)andforthecrucialpairUrb/Rbthe
numberof15linesisonlyjusthigherthanchancewouldexplain.
ItseemsthenthatanalysisbasedontheBarbilociwillbeveryeffective
atrevealingtheMart/TrivandAsh/Hampairs,butmuchlesseffectivefor
α and for Urb/Rb. The prominence of Mart/Triv in the Barbi loci is
likelytobenoaccident.TheTrivulziano1080manuscript–exceptional
fortheextraordinarybeautyofitsscript,withpageafterpagefreefrom
shoulddiscountthe49Ash/Hamlinesbythe10linesfoundinMart/Trivandthetwo
foundinα,hence12;anddiscountalsothe19 αlinesbythethreefoundinMart/Triv
andthetwofoundinAsh/Ham,hence14.
ThetextualtraditionofDante’sCommedia
23
anyblemish,andfamousforitsscribeFrancescodiSerNardo,itsage,the
consistencyofitslanguage(seethedescriptioninShaw)–wasalready
well-knowntoBarbi.ItisreasonabletoassumethatBarbi’sknowledgeof
Trivwasafactorinhischoiceofloci;hencethehighnumberofMart/Triv
linesfoundintheBarbilist.Accordingly,weshoulddiscountthenumbersoflinesgiveninthelastcolumnoftable4assuggestedinthefootnotetothetable,andusethenumbersgiveninbracketsasthebasefor
ourassessment.This(ofcourse)hasnoeffectontheefficacyoftheBarbi
listasfarasMart/Trivisconcerned,andtheAsh/Hamcount(37adjusted
variantsagainstanupperchanceboundof30)isalsostillbeyondwhat
couldbeexplainedbychance.ButthenumberofUrb/Rblinesonthe
Barbilistnowfallstosix:lowereventhantheexpectednumberwhich
simplechancewouldgive.
Sanguineti,aswehaveseen,basedhisentirestemmaticanalysison
theBarbiloci.Accordingly,hewouldhavecollatedjustsixlineswhich
(intheShawanalysis)alonecontainreadingslikelytohavebeenintroducedbythecommonancestorofUrb/Rb.Indeed,inhisanalysishe
referstoonlyfouroftheselines.17Shawisabletobaseherassertion
thatRbandUrbshareacommonancestorbelowthearchetypeonthe
308variantreadings,foundin297lines,identifiedbyVBaseworking
ontheresultsofthephylogeneticanalysis.Sanguineti,itappears,followingBarbi,examinedonlyahandfuloflineslikelytocontainreadingswhichevidencecommondescentofUrbandRbfromanancestor
belowthearchetype(Petrocchi’sβ).Hence,hisargumentthatUrband
Rbdonotshareanysuchancestor,thatUrbdescendsdirectlyfromthe
archetype,sharingnointermediateancestorwithanyotherextantmanuscript.ThisgivesUrb,inhisanalysis,extraordinaryanduniquestatus:
17
Thefourare:Inf.xi84menDiooffendeemenbiasimoaccatta;Inf.xiii63tanto
ch’i’neperde’lisonnie’polsi;Purg.iv72chemalnonseppecarreggiarFetòn;Purg.xvi
145.Cosìtornò,epiùnonvolleudirmi.ThetwoheappearsnottohaveusedareInf.xxvi
15rimontò’lducamioetrassemeeandPurg.v88IofuidiMontefeltro,iosonBonconte.AnothertwoappearinMart/TrivaswellasinUrb/Rb,andareusedbySanguineti:Inf.i28Poich’èiposatounpocoilcorpolasso;Purg.xviii57ede’primiappetibili
l’affetto;twoappearinMart/TrivbutarenotusedbySanguineti:Purg.xviii58chesono
invoisìcomestudioinape;Purg.xxiv61equalpiùagradireoltresimette;twoappearin
α andareusedbySanguineti:Inf.xvii50orcolceffoorcolpiè,quandosonmorsi;
Purg.xiii121tantoch’iovolsiinsùl’arditafaccia;threeappearinαandarenotusedby
Sanguineti:Inf.i116vedrailiantichispiritidolenti;Purg.xxiii2ficcavaïosìcomefar
suole;Par.i54efissiliocchialsoleoltrenostr’uso.ItshouldbenotedthatSanguineti
alsodoesnotuseanotherlineontheBarbilist,Par.I25,whichbothPetrocchiandShaw
regardassignificantforanalysisofUrb/Rb.
24
PeterRobinson
indeed,theauthorityofthisonemanuscriptUrbisequivalenttothat
ofalltheother800-plusmanuscriptsoftheCommediaputtogether.
Accordingly,SanguineticonsistentlyprefersthereadingsofUrbtothose
ofothermanuscripts,acceptingtheUrbreadingineverycaseexcept
wherethereisclearerror.
Sanguineti’sassertionoftheuniqueauthorityofUrbaffectshistext
significantly.Insome1406places,heacceptsthereadingofUrboverthat
foundinothermanuscriptsandacceptedbyPetrocchi.Ifyoureadthe
CommediainSanguineti’sedition,aboutonelineintenwillbedifferentfromthatgiveninthewidely-acceptedPetrocchitext(or,indeed,
fromanyeditionbasedonTrivorothermanuscriptsfromtheFlorentinetradition).AllthisfollowsfromhisassertionthatUrbisnotgeneticallyrelatedtoRb,whichitselfcanbetracedtohisapparentexaminationofjustahandfulofthemanyvariantswhichinShaw’sanalysis(as
inPetrocchi’s)evidencethisrelationship.
Howdidthissituationarise?ItappearsthatSanguinetiputallhistrust
intheBarbiloci,asthebasisforacompleteaccountoftheentiretradition.However,theBarbilociincludeveryfewofthevariantsidentified
bytheVBaseanalysisasevidenceforthegeneticrelationshipofUrband
Rb.Accordingly,hefailedtoseethisrelationship,andthisleddirectlyto
hishypothesisoftheuniqueauthorityofUrb.Itcanbearguedthatthis
failureisexacerbatedbecauseitappearsSanguinetididnotuseallthe
Barbiloci:heusedonlyfourofthesixlinesgivenbyBarbiwhichinour
analysisevidencetheUrb/Rbrelationshipalone,andalsousedonlyfourof
theninelinesgivenbyBarbiwhichevidenceUrb/Rbandalsoeither
orbothofMart/Trivandα.TheBarbilocicontainatotaloffifteenlines
whichevidenceUrb/Rb,eitheronitsown(sixlines)orincombination
withothergroups(ninelines);Sanguinetiusesonlyeightofthese.One
isremindedthatkeypartsofSanguineti’sargumentrestonaveryfew
readings:heusesjustfourtoeliminatemorethanfourhundredmanuscriptsfromconsideration(seeShaw’sdiscussion,«Introduction–Sanguineti:TheTradition»).Yet,evenifSanguinetihadusedalltheUrb/
RbvariantsincludedintheBarbiloci,itisstilllikelythathewouldhave
reachedthesameconclusion.
Indeed,PetrocchiwarnedexplicitlyagainstuseoftheBarbilociasa
baseforthetextualreconstructionofthewholetradition.Inhis«Proposteperuntesto-basedellaDivinaCommedia»,FilologiaRomanza,
II(1955),pp.337-365,hearguesthatwhileBarbi’shypothesis(thata
geneticreconstructionofthetraditionispossibleonthebasisofasmall
numberofchosenlines)wasplausiblein1890,researchintothetradi-
ThetextualtraditionofDante’sCommedia
25
tionintheyearssinceBarbi’sarticleshowedthatthecomplexitiesinthe
traditionareofsuchanaturethatanysuchpartialanalysiscanyieldonly
«unprimoegenericoorientamentonelfittodeirapportitraicodici»:
«apreliminaryandgeneralsenseofdirectionwithinthethicketofrelationshipsamongthemanuscripts».Theanalysisinthisarticlesuggests
thatPetrocchiisexactlyright.AnalysisbasedontheBarbilocishowsthe
broadestfeaturesofthetradition:thustheclearlymarkedpairsAsh/Ham
andMart/Triv.Butitfailstoshowmoreelusive,yetreal,relationships,
aswehaveseenforUrb/Rb.Suchmapsaredangerous.Aswenotedearlier,Petrocchiaccordinglytookaquitedifferentdirection.Ratherthan
sampleafewreadingsacrossverymanymanuscripts,asBarbisuggests,
hechosetolimitthenumberofmanuscriptssurveyedandexaminedall
thereadingsinthosemanuscripts.TheShaweditionfollowsthispath,
narrowingthenumberofmanuscriptsstillfurthertojustseven(includingone,LauSC,notamongtheanticavulgatamanuscriptsusedbyPetrocchi),andusinganarrayofanalytictechniquesnotavailabletoPetrocchi.ThecloseagreementbetweenShawandPetrocchionalmostallpoints
maybeseenasavindicationofPetrocchi’smethod.18
ItcouldbearguedthatitisnosurprisethatSanguinetifailedtocreate
anadequateaccountofthemanuscriptrelationsonthebasisoftheBarbi
loci.Onemightquestion(asPetrocchidid)thefundamentalmethodology,oftryingtocreateamanuscriptstemmaofsovastatradition,and
solongatext,onasampleoflessthan3%ofthematerial.Onemightalso
pointtoweaknessesinSanguineti’suseoftheloci:hisdependenceona
veryfewreadingstoclassifyverymanymanuscripts;hisapparentfailuretousealltheloci.Butthereisoneaspectofthiscomparisonofthe
ShawresultswiththeBarbilociwhichmightbesurprising.Inotedabove
thatBarbidesignated396ofthe14233linesoftheCommediaasmost
likelytobeusefulfortracingthetextualfamiliestotheoriginal(«rintracciarelefamigliedeitestiapenna»).TheShaw/Robinsonanalysisidentified1885linesascontainingvariantsdistinctiveoffamilygroupings.
OnemightexpectthatahighproportionofBarbi’s396lineswouldbe
amongthe1885identifiedintheShawanalysis:infact,thereare121lines
incommon.Whilethisisfarmorethansimplechancemightpredict(as
explainedabove),itseemslowerthanonemightexpect.BarbihasanearlegendarystatusamongItalianphilologists.Hewasinhisearly20satthe
Butnotonallpoints:PetrocchiarguesthatHamisadescendantofAsh;Shaw
showsthatAshandHamarebothindependentlydescendedfromthesameancestor
withinthetradition.
18
26
PeterRobinson
timeofpublicationofthelociarticle,atthebeginningofanillustrious
careeraseditor,philologistandDantescholar,inwhichhecametoknow
thebusinessofeditingandthemanuscriptsandtextoftheCommedia
asfewhaveeverdoneorwilldo.Ifoneweretochooseanyoneperson
toselectagroupoflinesmostlikelytobeusefulforstemmaticanalysis,Barbiwouldbethechoiceofmanyscholars.Yet,lessthanathirdof
thelinesheselectedappearinthelistof1185linesidentifiedbytheShaw
analysis.Thatis:sometwothirdsofthelineshechose(275of396)turned
outnottobeproductiveforShaw’sanalysis.Onecouldmakesomeallowance,thatsomeofthese275linesmightberevelatoryofrelationsamong
manuscriptsnotstudiedbyShaw,butthefundamentalpositionofsixof
thesevenShawmanuscriptsinPetrocchi’saccountoftheearlyhistory
ofthetradition(theareaofmostconcerntoBarbi,asforanyeditor)make
thisadifficultargumenttosustain.19
Why,then,doesBarbimanagetoidentifysofewlineswhichareactuallyproductiveforphylogeneticanalysis?Barbi’saccountoftheprinciplesonwhichhechosetheselinesishighlycompressed,justafew
sentences.Heremarksthatitwouldbe«inutileaffatto»tosetoutthe
reasonsforthechoiceversebyverse,andstatesthegeneralprinciple
19
Petrocchidividesthemanuscriptsoftheanticavulgataintotwomajorgroups:α
andβ.Hedividesβ intotwogroups,dande,withRbandUrbastwoofthethreerepresentativesofe.Hedividesαintothreegroupings:abc.MartandTrivarethesolerepresentativesofa;AshandHamrepresenttwo(inShaw’sanalysis;PetrocchiseesHam
asdescendedfromAsh)ofsomefivebranchesofb;theoriginaltextofMart(«M0»in
Shaw)isclosetoVat,aprimaryrepresentativeofPetrocchi’sc.Thus,bothprimaryand
allthreesecondarybranchesofthetraditiondefinedbyPetrocchiarerepresentedinthe
sixmanuscriptsAshHamMartTrivUrbRbstudiedbyShaw.TheonlybranchofPetrocchi’sstemma(ifitisabranch)notrepresentedisthatoftheanomalousLandinomanuscript,whichPetrocchiplacesasablendofreadingsdrawnfromc(via«c1»)andβ(via
«d»).ItshouldbenotedthatShaw’sstudyalsoshowsthattheLaurenzianoSantaCroce
manuscript,muchfavouredbyearlyeditorsoftheCommedia(Wittethoughtitthe
bestsurvivingmanuscriptandusuallychoseitsreadingsaheadofthoseofothermanuscriptsforhis1862edition;MooreandCasellaalsoratedithighly;seeShaw«Introduction»,notes31and47)andchosenbySanguinetiasoneofhisseven,isactuallyaclassicallycontaminatedmanuscript,combiningreadingsinitsoriginaltextdrawnfromfour
ofthefivesub-archetypesidentifiedbyPetrocchi(thatis:abce),andthecontamination
iscompoundedbythecorrectionsintroducedintothemanuscript,withthe«c2»correctorapparentlyintroducingreadingsfromanotherwitnesswhichappearstocombinereadingsfromPetrocchi’scandebranches(RobinsonandShaw«Phylogenetic
Analysis»,inShawCommedia.)ThisanalysisofLauSCaccordinglyconfirmsPetrocchi’sargument:thatmanuscriptswrittenafter1355arelikelytocontainsuchmixesof
readingsdrawnfrommultiplebranchesastobeofnovaluetowardsreconstruction
oftheearliestphasesofthetradition.
ThetextualtraditionofDante’sCommedia
27
guidingthechoice:thereshouldbetwoormorevariantspresent,each
inseveralmanuscripts.Beyondthis,hesetsoutthereadingswhichhe
excluded:readingswhichshowtranscribersandcommentatorsexhibitingcriticaljudgement;andthosewhereseparatescribesarehighlylikely
tointroducethesamereadingforreasonsof«eufonia,sinonimia,correlazionegrammaticale»,andnotesthathepaidspecialattentiontothe
helpwhichmajorerrorscanaffordtowardstheclassificationofmanuscripts(«Abbiamoanchepensatoall’aiutocheporgonoperlaclassificazionedeimanoscritti,glierrorigrossolanidilezione»).
Wemaycomparethesecriteriawithsomeofthereadingsidentified
bytheShaw/Robinsonanalysisasindicativeoffamilyrelations.Here
arefiveofthe308readingsidentifiedaslikelytohavebeenintroduced
bythecommonancestorofUrb/Rb;noneofthesefivelinesappear
amongtheBarbiloci.
Inf.i89:aiutamidalei,famososaggio,
famosoesaggio
LauSCRbUrbFS
famososaggio
AshHamMartTrivPET
Inf.ii71:vegnodellocoovetornardisio;
di
LauSCRbUrbFS
del
AshMartTrivPET
dal
Ham
Inf.ii110:afarlorprooafuggirlordanno,
proea
Mart-origRbUrbFS
pronea
AshLauSC-c2Triv
prodeeta
Ham
pro[..]a
LauSC-orig
prooa
Mart-c2PET
Inf.iii3:permesivatralaperdutagente.
nela
RbUrbFS
trala
AshHamLauSCMartTriv-c1PET
tra
Triv-orig
Inf.iii22:Quivisospiri,piantiealtiguai
altri
Ash-origRbUrb
alti
Ash-c2HamLauSCMartTrivFSPET
28
PeterRobinson
WecanseewhyBarbiwouldnothavechosenthesereadings.Allhave
theslightcharacterofvariantswhichmighthaveappearedreadilyatany
pointofthetradition,andhencearelikely(tousethetermemployed
byPetrocchiandShaw)tobe«polygenetic»;inShaw’sdefinition«an
errorliabletoariseindependentlyinindependentmanuscripts».One
canimaginethatnotextualscholar,educatedtorejecterrorslikelytobe
polygeneticasindicativeofanything,wouldselecttheseasthebasisfor
anykindofargument:indeed,onlyone,trala/nelaInfiii.3,isusedby
ShawinheraccountofRb/Urb.YettheShaw/Robinsonanalysisofthe
distributionofthevariantsacross94779distinctsitesofreadings(usually,oneword,butoftenaphrase)overthe14233linesoftheCommediasuggeststhattheappearanceofthesefivereadingsinUrb/Rbresults
fromtheirintroductionintoanancestorcopysharedbyUrbandRb.
Thattwoofthesefive(Inf.i89andii71)appearalsoinLauSCislikely
tobetheresultofcontaminationinthatmanuscript(seefootnote20).
ThisleavesonlyInf.ii110«proea»intheAldinetextwhichisthebase
oftheMartinicollation(itselfderivedfromVat)andInf.iii22«altri»in
theoriginaltextofAshaslikelypolygeneticerrors,ifnotderivingfrom
contamination.
Tosumup:ahighproportionofthe1885linesidentifiedbytheShaw/
RobinsonanalysisasindicativeofmanuscriptrelationscontainvariantswhichwouldnotsatisfyBarbi’scriteriaasbeingunlikelyapriori
tohaveariseninindependentcopies.Yet,thesesamevariantsareuseful
forouranalysis,preciselybecauseasagrouptheydidnotariseinindependentcopies.Thekeyphrasehereis«asagroup»:individualvariantsamongthem(say,amongthe308Urb/Rbvariants)didappearin
othermanuscripts,sometimesperhapsbycontamination,sometimes
bypolygenesis.Butothermanuscriptsoutsideeachgroupappearalongsidethemembersofthegroupinascattered,randomfashion,aswesee
inthesefivevariants:Urb/Rbagreeinallfive,oncewithnoothermanuscript(iii3),twicewithLauSC(I89,ii71),oncewiththeoriginaltext
ofMart(ii110),oncewiththeoriginaltextofAsh(iii22).Thisisexactly
asweshouldexpectinarealmanuscripttradition:thatthescribeof
thecommonancestorofUrb/Rbintroducedsome308readingsintothe
copy,whichthendescendedintoUrb/Rb.Ofthosereadings,weshould
expectafewofthemtoappearinothermanuscripts,bycontamina-
tion or by polygenesis: thus, we find 40 of the 308 in Ash (also, 2
Ash-orig),11inLauSC(11LauSC-orig,4LauSC-c1,5LauSC-c2,4in
Mart(Mart-orig13),9inTriv,46inHam,whileall308appearinboth
UrbandRb.
ThetextualtraditionofDante’sCommedia
29
Thissuggeststhatonecannotdeterminewhetheraparticularvariantismonogenetic(as,ideally,wewouldwantallvariantstobe)or
polygeneticpurelyonthebasisofthevariantitself,asBarbisuggests
wedo.Onemustalsotakeintoaccountboththeactualdistributionof
manuscriptswhichhavethevariant,andtheconsistencyofthismanuscriptdistributionacrossmanyothervariants.Whenweseethepair
Urb/Rbappearoverandover,sometimesontheirown,sometimeswith
avaryingcastofothermanuscripts,thenweareentitledtopresumethat
thisconsistencyofdistributionitselfmaysuggestageneticrelationship
betweenthemanuscripts.AlthoughbothPetrocchiandShawpayclose
attentiontovariantswhichtheyregardaslikelytobemonogenetic,on
thebasisofthevariantitself(asBarbidoes),theyalsoareawareofthe
significanceofaconsistentpatternofagreements,foundinvariantafter
variant.ShawcitesPetrocchi’sargumentthatthetrioUrb/Rb/Madhave
acloseness«nettamentesuperioreperfoltezzadistatisticaenaturadei
casiaqualsiasiaffinitàtrailtestimoneurbinateeglialtri»(Introduction,
citingPetrocchiIntroduzione334-389),andpointsoutthesignificance
ofthephrase«foltezzadistatistica»:«statisticaldensity».AsInotedearlier,ShawexpandsonPetrocchi’sperception,asshediscussesanumber
ofreadingssupportingthegeneticrelationshipofUrb/Rbwhich,she
concedes,areeachindividuallypossiblypolygenetic,butwhichtaken
togetherevidenceageneticrelationship.Thusherconclusion:
Theseareofcoursesmallerrors,possiblypolygenetic.Eachoneinitselfwould
countforverylittle.Butitisthecumulativepicturebuiltupofawholeseries
ofsuchsmallerrorsrightacrossthetextandsharedwithnoorveryfewother
manuscriptswhichissignificantandsuggestsacommonexemplar.
Seenthisway,PetrocchiandShaw’suseof«foltezzadistatistica»alongsideconsiderationofthemonogeneticorpolygeneticcharacterofeach
variantisanextensionofBarbi’smethod,notarepudiationofit.Barbi,
PetrocchiandShawallagreethatitisunsafetorelyoncertainindividual
variants,consideredontheirown,asindicativeofrelationsacrossthewhole
tradition.Indeed,ShawcriticizesSanguineti(asInoteabove)forbasing
hisanalysisonafewvariants.However,bothPetrocchiandShawareable
touse«foltezzadistatistica»becausetheydotakeintoaccounteveryvariant,everyreading,inthemanuscriptstheysurvey–anapproachonlypossiblebecausetheychoosenottobasetheiranalysisonasampleofreadings
fromagreatmanymanuscripts(theBarbilocimethodology,followedby
Sanguineti)butonallthereadingsfromasmallnumberofmanuscripts.
30
PeterRobinson
Barbi;Petrocchi;Shaw:morethanacenturyofDantetextualscholarship,acenturywhichsawItaliantextualscholarshipabsorb,andthen
reshapeitselfinreactionto,theimpactofLachmannianmethodsand
theoppositiontothesemethodsledbyBédierandothers;ascholarlytraditionwhichisnowreckoningwiththefirstimpactofanotherrevolution,asdigitalmethodsopenupnewwaysofapproachinglargetextualtraditions.Weshouldmentiontwomorenames:GiogioPasquali,
whoseStoriadellatradizioneecriticadeltesto(Florence,LeMonnier1934)
repudiatedtherigidmethodologyoutlinedinMaas’sTextkritik(Leipzig,
Teubner1927),withoutrejecting,asBédierdid,thepossibilityofmaking
senseoflargeandcomplextraditions;andGianfrancoContini.In1935,
whenContinireviewedPasquali’sbook,hewasaroundthesameageas
Barbiwhenhewascompilingthelistofloci.Barbi’sownLanuovafilologiael’edizionedeinostriscrittoridaDantealManzoni(Florence:Sansoni1938)followsPasqualiinadvocatingwhatPugliattidescribesasa
«virtuousmiddleway»betweenLachmannandBédier.20Continifollows
this«middleway»,assertingthroughouthiscareerthatintelligentuseof
everyavailablemethod,includingstemmaticswhereappropriate,joined
withdeepknowledgeofthemanuscriptsandthetextstheycarry,would
permitausefulunderstandingofwholemanuscripttraditions.Shaw
studiedwithContini,firstasapost-graduatefrom1962to1964,andthen
from1966asshewroteherDott.Lett.dissertation,aneditionoftheearliestItaliantranslationoftheMonarchia,withContiniashersupervisor,
thusbracketingtheappearanceofPetrocchi’sgreateditionin1966.With
Contini’sencouragement,shetheneditedMarsilioFicino’stranslationof
theMonarchia.AgainwithContini’sencouragement,in1981shewrote
anarticlecriticalofsomeaspectsofRicci’seditionoftheMonarchia.21
Afterthearticlewaspublished,ContinisuggestedtoShawthatsheundertakeaneditionoftheMonarchia.Shawhesitatedatfirst:shewasaware
thatafactorintheproblemsofRicci’seditionoftheMonarchiawasthe
amountofdatageneratedbyatextinmanyversions,andRicci’sinability
tomanageallthisdata.Itwasonlywhensheundertookanintroductory
courseincomputingattheCambridgeUniversityComputingCentrein
October1985andsawthepossibilityofusingacomputertorecordand
20
PaolaPugliatti,«TextualPerspectivesinItaly:fromPasquali’sHistoricismtothe
Challengeof“Variantistica”(andBeyond)»,Text,11(1998),pp.155-188[p.162].
21
Shaw’sdoctoraldissertationwaspublishedas«Ilvolgarizzamentoineditodella
Monarchia»,Studidanteschi,XLVII(1970),pp.59-224;theeditionoftheFicinotranslationas«LaversioneficinianadellaMonarchia»,Studidanteschi,LI(1978),pp.289-407;the
articleonRicciin«SultestodellaMonarchia»,Studidanteschi,LIII(1981),pp.187-217.
ThetextualtraditionofDante’sCommedia
31
exploreallthedatageneratedbyfulltranscriptsofeachmanuscriptthat
shecommittedherselftomakingtheedition–and,tousingcomputer
methodstodoso.22
ShawdiscussedthisnewmethodologywithContini,andheencouragedherinthispath.ShedescribedherworkinapaperattheconferenceinFlorencein1988,convenedtocelebratethecentenaryofthe
foundingoftheSocietàDantescaItalianain1888.23Herdecision,tobase
theeditiononfull-texttranscriptsofthemanuscriptslooksbackto
Petrocchi,justasherdecisionthatthetranscriptsbemadeinmachinereadableformlooksforwardtothedevelopmentofcomputer-based
systemsforcomparisonandanalysis.In1991,Shawhadcompletedthe
firsttranscriptsoftheMonarchiamanuscriptsandwasexploringhow
thesetranscriptsmightbeusedtomakeanedition.24Iwasstartingwork
onwhatbecametheCollatesuiteofsoftwaretoolsforcreatingeditions
frommultipletranscriptsofworks,andalsodiscoveringhowtoolsfrom
evolutionarybiologycouldbeusedtoadvanceunderstandingoflarge
textualtraditions.Shawcametoseeme(itisafortunatechancethather
dentistwasbasedinOxford,whereIwasthenworking),andthisbegan
atwo-decadecollaborationwhichissuedinhereditionsoftheMonarchiaandCommedia.
Barbi,Pasquali,Petrocchi,Contini,Shaw.Thesescholarsarejoined
inacommonbelief:sensecanbemadeoflargemanuscripttraditions
usinggoodmethodandexpertphilologicalknowledge,andusefulediAsanindexofthedifficultyofthiswork,inthosedaysbeforecomputersbecame
portableandpersonal:Shawtranscribedthemanuscriptsontopaperfromamicrofilm
readerattheUniversitylibrary;thencarriedthetranscriptsacrosstheCamtothecomputingcentre,whensheenteredthemintotheCambridgemainframeusingthe«Zed»
lineeditor.
23
Published,includingsomeprintoutsofhertranscripts,in«Perunnuovotestocritico
dellaMonarchia»,inLaSocietàDantescaItaliana1888-1988.AttidelConvegnoInternazionalediStudiDanteschitenutoaFirenzenelnovembre1988,Firenze,1995,pp.435-444.
24
Otherscholarsatthistimewerecomingtotheviewthatfull-textcomputer-readable transcripts were the way forward for textual scholarship applied to large manuscript traditions. In particular, Francesco Mazzoni, Contini’s successor as President
oftheSocietàDantescaItaliana,encouragedhisstudentstomakefull-texttranscriptsof
Commediamanuscripts,anddevelopedasetoftranscriptionguidelinesforthiswork.
These«principiditrascrizione»maybefoundathttp://www.danteonline.it/italiano/
risorse.htm(withanEnglishtranslationbyShaw).Transcriptionsofseveralmanuscriptsareavailableonthedanteonlinesite.Mazzonihimself,andotherassociatesof
theSDI(particularlyPaulaLaurella),wereextremelysupportiveofShaw’scontinuing
workontheMonarchia(herprinteditionwaspublishedintheSDI’sEdizioneNazionale
series;theSDIco-publishedtheDVDversion)andontheCommedia.
22
32
PeterRobinson
tionsofworkscanbefoundedonthatsense.Full-textmachinereadabletranscriptsofwholemanuscripts,passedthroughanarrayofcomputer-basedmethods,arenowpartofaphilologist’sarmory,tobeused
inconjunctionwith(andnotinoppositionto)knowledgeofthemanuscripts,ofthetextitselfanditscontexts,andwitheditorialintuition.
Thisseemsausefulstarting-pointforfurtherexplorationsintotheCom-
mediatradition.
UniversityofSaskatchewan