On Dating Proto-Indoeuropean Via Bayesian Phylogenetic Inference

On dating Proto-Indoeuropean via Bayesian phylogenetic inference

Igor Yanovich1, Armin Buch, Johannes Dellert, Marisa Köllner,Fabr´ıcio Marcel Ferraz Gerardi, Roland Mühlenbernd, Johannes Wahle and Gerhard Jäger2

Institute of Linguistics, T¨ubingenUniversity 1Carnegie Mellon 2Netherlands Institute for Advanced Study

MPI for the Science of Human History, Jena

October 13, 2015

J¨ager,Yanovich et al. (T¨u) On phylogenetic dates MPI Jena 1 / 38 Phylogenetic age constraints

Jäger,Yanovich et al. (Tü) On phylogenetic dates MPI Jena 2 / 38 Bouckaert et al: based (mostly) on written record this study upper and lower limit: derive upper and lower confined to last two limits from archaeological millenia findings only lower limit: go up to last four millenia

J¨ager,Yanovich et al. (T¨u) On phylogenetic dates MPI Jena 3 / 38 Proto-Indo-Iranian

Indo-Iranian Identiﬁcation of Proto-II with Andronovo culture (Kuz’ima 2007) accepted by (some) proponents both of Anatolian and of Steppe theory time constraint: 1900–1600 BCE “Following the dissolution of the Balkan PIE linguistic area it seems likely that there was some further and signiﬁcant punctuation episode in the Pontic steppe area which motivated the eastern dispersal of early Indo-Iranian into areas where the Early Steppe form of (Proto) IE had already been spoken for some time. Early Indo-Iranian is often correlated with the Andronovo culture of the Eurasian steppes, seen after 2000 BC (Kuzmina 1994)” Renfrew 1999

J¨ager,Yanovich et al. (T¨u) On phylogenetic dates MPI Jena 4 / 38 The Tocharian split

Tocharian Steppe theory (Anthony): migration from the steppe towards the Altai mountain ca. 3300–3000 BCE Anatolian theory (Renfrew 1999): eastward migration from the Balkans into the steppes archaeological evidence for intense cultural exchange between steppes and the Balkans by the end of the 4th millenium if identiﬁed with Tocharian split, this also gives a time constraint ca. 3300–3000 BCE

J¨ager,Yanovich et al. (T¨u) On phylogenetic dates MPI Jena 5 / 38 Phylogenetic age constraints

J¨ager,Yanovich et al. (T¨u) On phylogenetic dates MPI Jena 6 / 38 Replication of Bouckaert et al

175

150

125

100

Frequency 75

0 5000 6000 7000 8000 9000 10000 11000 12000 13000 treeModel.rootHeight mean: 7685 BP; 95% HDI: 6053–9545

J¨ager,Yanovich et al. (T¨u) On phylogenetic dates MPI Jena 7 / 38 Adding Indo-Iranian constraint

175

150

125

100

Frequency 75

0 5000 6000 7000 8000 9000 10000 11000 12000 13000 14000 15000 treeModel.rootHeight mean: 7162 BP; 95% HDI: 5525–8931

J¨ager,Yanovich et al. (T¨u) On phylogenetic dates MPI Jena 8 / 38 Adding Tocharian constraint

200

175

150

125

100 Frequency

0 5000 6000 7000 8000 9000 10000 11000 12000 13000 treeModel.rootHeight mean: 7295 BP; 95% HDI: 5316–8982

J¨ager,Yanovich et al. (T¨u) On phylogenetic dates MPI Jena 9 / 38 Adding both constraints

250

200

150 Frequency

100

0 5000 6000 7000 8000 9000 10000 11000 12000 13000 14000 treeModel.rootHeight mean: 7139 BP; 95% HDI: 5782–8622

J¨ager,Yanovich et al. (T¨u) On phylogenetic dates MPI Jena 10 / 38 Adding both constraints

Lycian Luvian Hittite Wakhi Iron_Ossetic Digor_Ossetic Avestan Old_Persian Afghan Waziri Tadzik Persian_List Kurdish Baluchi Romani Singhalese Urdu Lahnda Hindi Sindhi Marwari Gujarati Marathi Bengali Oriya Assamese Bihari Nepali_List Kashmiri Vedic_Sanskrit Oscan Umbrian Catalan Spanish Portuguese_ST French Walloon Provencal Friulian Italian Romansh Ladin Sardinian_L Sardinian_C Sardinian_N Romanian_List Vlach Latin Breton_ST Breton_SE Breton_List Cornish Welsh_N Welsh_C Irish_A Scots_Gaelic Old_Irish Gothic Old_English English_ST Old_High_German Luxembourgish German_ST Frisian Flemish Dutch_List Old_Norse Faroese Icelandic_ST Riksmal Danish Swedish_VL Swedish_Up Swedish_List Old_Church_Slavonic Serbocroatian Macedonian Bulgarian Slovenian Russian Polish Ukrainian Byelorussian Czech Slovak Czech_E Lusatian_L Lusatian_U Latvian Lithuanian_ST Old_Prussian Armenian_List Armenian_Mod Classical_Armenian Albanian_Top Albanian_G Albanian_K Albanian_C Ancient_Greek Greek_ML Greek_Mod Tocharian_B Tocharian_A

9000 8000 7000 6000 5000 4000 3000 2000 1000 0

J¨ager,Yanovich et al. (T¨u) On phylogenetic dates MPI Jena 11 / 38 Summary

10000

9000

8000

7000 treeModel.rootHeight

6000

5000 IE2011_RelaxedCovarion_AllSingletonsGeo.logtreeModel.rootHeight IE2011_InIr.logtreeModel.rootHeight IE2011_Tocharian.logtreeModel.rootHeight IE2011_InIr_Tocharian.logtreeModel.rootHeight

Bayes factor Steppe Theory vs. Anatolian Theory goes up from 1:1 to 4:1

Jäger,Yanovich et al. (Tü) On phylogenetic dates MPI Jena 12 / 38 And now for something completely different...

J¨ager,Yanovich et al. (T¨u) On phylogenetic dates MPI Jena 13 / 38 Borrowings in IELex

general policy: loanwords are manually identiﬁed and marked as such usually excluded in phylogenetic analyses typical example: mountain (borrowed into English from French) sometimes decision is controversial: Russian sobaka ‘dog’ is assigned to the same cognate class as German Hund, but there is a debate in the literature whether it is an Iranian or a Turkic loan (it is certainly not inherited) deep borrowings might not be recognizable as such

Impact on phylogenetic inference?

J¨ager,Yanovich et al. (T¨u) On phylogenetic dates MPI Jena 14 / 38 Semi-automatic loanword detection

In an ideal phylogeny, each cognate class should emerge exactly once (it might get lost several times). ⇒ Dollo-model of character evolution parallel changes 0 → 1 possibly indicative for borrowing There might be other reasons, such as parallel semantic change wrong phylogeny incomplete sampling of synonyms ... this talk: Maximum likelihood character state reconstruction manual identiﬁcation of parallel changes

J¨ager,Yanovich et al. (T¨u) On phylogenetic dates MPI Jena 15 / 38 Semi-automatic loanword detection

Hittite ● Luvian ● Tocharian B ●Tocharian A ●Classical Armenian ● Armenian List ● Armenian Mod Ancient Greek ● ● Greek K ● Greek Mod ● Tsakonian ● Greek Md ● Greek D ● Greek Ml Gothic ● ● Old English ●Old High German ● Frisian ● Afrikaans ● Dutch List ● Flemish ● German ● Standard German Munich ● Pennsylvania Dutch ● Letzebuergesch ● Schwyzerduetsch Old Gutnish ● ● Old Swedish ● Faroese ● Icelandic St Old Norse ● ● Norwegian ● Stavangersk ● Danish Fjolde ● Danish ● Oevdalian ● Gutnish Lau ● Swedish ● Swedish Vl ● Swedish Up ● Gaulish Old Irish ● ● Manx ● Gaelic Scots ● Irish B ● Irish A Old Cornish ● ● Old Welsh ● Welsh N ● Welsh C ● Cornish ● Breton St ● Breton List ● Breton Se Latin ● ● Rumanian List ● Vlach ● Sardinian L ● Sardinian N ● Sardinian C ● Italian ● Dolomite Ladino ● Friulian ● Ladin ● Romansh ● Provencal ● French ● Walloon ● Catalan ● Portuguese St ● Brazilian Old Prussian ● ● Latvian ● Lithuanian St ● Lithuanian O Old Church Slavonic ● ● Slovenian ● Serbocroatian P ● Serbian ● Serbocroatian ● Macedonian P ● Macedonian ● Bulgarian ● Bulgarian P ● Slovenian P ● Russian P ● Russian ● Ukrainian P ● Byelorussian P ● Byelorussian ● Ukrainian ● Polish ● Polish P ● Lower Sorbian ● Upper Sorbian ● Czech P ● Slovak P ● Czech ● Czech E ● Slovak ● Prasun ● Kati ● Ashkun Avestan ● ● Sogdian ● Ossetic ● Iron Ossetic ● Digor Ossetic ● Wakhi ● Sariqoli ● Shughni ● Waziri ● Pashto ● Persian ● Tadzik ● Baluchi ● Zazaki ● Kurdish Vedic Sanskrit ● ● Kashmiri ● Singhalese ● Gypsy Gk ● Khaskura ● Nepali ● Bihari ● Bengali ● Oriya ● Assamese ● Marathi ● Gujarati ● Marwari ● Sindhi ● Magahi ● Hindi ● Urdu ● Panjabi St dog:A ● Lahnda

J¨ager,Yanovich et al. (T¨u) On phylogenetic dates MPI Jena 16 / 38 Semi-automatic loanword detection

J¨ager,Yanovich et al. (T¨u) On phylogenetic dates MPI Jena 17 / 38 Semi-automatic loanword detection

J¨ager,Yanovich et al. (T¨u) On phylogenetic dates MPI Jena 18 / 38 Semi-automatic loanword detection

J¨ager,Yanovich et al. (T¨u) On phylogenetic dates MPI Jena 19 / 38 Semi-automatic loanword detection

J¨ager,Yanovich et al. (T¨u) On phylogenetic dates MPI Jena 20 / 38 Semi-automatic loanword detection

J¨ager,Yanovich et al. (T¨u) On phylogenetic dates MPI Jena 21 / 38 Semi-automatic loanword detection

● Hittite ● Luvian ● Tocharian B ● Tocharian A ● Classical Armenian ● Armenian List ● Armenian Mod ● Ancient Greek ● Greek K ● Greek Mod ● Tsakonian ● Greek Md ● Greek D ● Greek Ml ● Gothic ● English ● Old English ● Old High German ● Frisian ● Afrikaans ● Dutch List ● Flemish ● German ● Standard German Munich ● Pennsylvania Dutch ● Letzebuergesch ● Schwyzerduetsch ● Old Swedish ● Faroese ● Icelandic St ● Old Norse ● Norwegian ● Stavangersk ● Danish Fjolde ● Danish ● Oevdalian ● Gutnish Lau ● Swedish ● Swedish Vl ● Swedish Up ● Gaulish ● Old Irish ● Manx ● Gaelic Scots ● Irish B ● Irish A ● Old Cornish ● Old Breton ● Welsh N ● Welsh C ● Cornish ● Breton St ● Breton List ● Breton Se ● Latin ● Rumanian List ● Vlach ● Sardinian L ● Sardinian N ● Sardinian C ● Italian ● Dolomite Ladino ● Friulian ● Ladin ● Romansh ● Provencal ● French ● Walloon ● Catalan ● Spanish ● Portuguese St ● Brazilian ● Old Prussian ● Latvian ● Lithuanian St ● Lithuanian O ● Old Church Slavonic ● Slovenian ● Serbocroatian P ● Serbian ● Serbocroatian ● Macedonian P ● Macedonian ● Bulgarian ● Bulgarian P ● Slovenian P ● Russian P ● Russian ● Ukrainian P ● Byelorussian P ● Byelorussian ● Ukrainian ● Polish ● Polish P ● Lower Sorbian ● Upper Sorbian ● Czech P ● Slovak P ● Czech ● Slovak ● Prasun ● Kati ● Ashkun ● Avestan ● Sogdian ● Ossetic ● Iron Ossetic ● Digor Ossetic ● Wakhi ● Sariqoli ● Shughni ● Waziri ● Pashto ● Persian ● Tadzik ● Zazaki ● Vedic Sanskrit ● Kashmiri ● Gypsy Gk ● Khaskura ● Nepali ● Bihari ● Bengali ● Oriya ● Assamese ● Marathi ● Gujarati ● Marwari ● Sindhi ● Magahi ● Hindi ● Urdu ● Panjabi St mouth:A ● Lahnda

J¨ager,Yanovich et al. (T¨u) On phylogenetic dates MPI Jena 22 / 38 Semi-automatic loanword detection

J¨ager,Yanovich et al. (T¨u) On phylogenetic dates MPI Jena 23 / 38 Semi-automatic loanword detection

J¨ager,Yanovich et al. (T¨u) On phylogenetic dates MPI Jena 24 / 38 Semi-automatic loanword detection

no borrowing: reﬂexes with diﬀerent meanings in other branches

J¨ager,Yanovich et al. (T¨u) On phylogenetic dates MPI Jena 25 / 38 Semi-automatic loanword detection

J¨ager,Yanovich et al. (T¨u) On phylogenetic dates MPI Jena 26 / 38 Semi-automatic loanword detection

J¨ager,Yanovich et al. (T¨u) On phylogenetic dates MPI Jena 27 / 38 Semi-automatic loanword detection

candidate for borrowing from Romance into Celtic

J¨ager,Yanovich et al. (T¨u) On phylogenetic dates MPI Jena 28 / 38 Semi-automatic loanword detection

J¨ager,Yanovich et al. (T¨u) On phylogenetic dates MPI Jena 29 / 38 Semi-automatic loanword detection

J¨ager,Yanovich et al. (T¨u) On phylogenetic dates MPI Jena 30 / 38 Semi-automatic loanword detection

J¨ager,Yanovich et al. (T¨u) On phylogenetic dates MPI Jena 31 / 38 Semi-automatic loanword detection

J¨ager,Yanovich et al. (T¨u) On phylogenetic dates MPI Jena 32 / 38 Semi-automatic loanword detection

J¨ager,Yanovich et al. (T¨u) On phylogenetic dates MPI Jena 33 / 38 Semi-automatic loanword detection

J¨ager,Yanovich et al. (T¨u) On phylogenetic dates MPI Jena 34 / 38 Semi-automatic loanword detection

J¨ager,Yanovich et al. (T¨u) On phylogenetic dates MPI Jena 35 / 38 Semi-automatic loanword detection

J¨ager,Yanovich et al. (T¨u) On phylogenetic dates MPI Jena 36 / 38 Semi-automatic loanword detection

in total 109 out of 2,496 (non-trivial) characters where manually identiﬁed as probable loans Bayesian phylogenetic analysis, using the constraints from Bouckaert et al., with full and with “cleaned” data set using all 150 varieties from IELex using diﬀerent software and tree prior (BayesPhylogeny, Yule tree prior)

J¨ager,Yanovich et al. (T¨u) On phylogenetic dates MPI Jena 37 / 38 mean = 6440 mean = 6460

95% HDI 95% HDI 5500 7510 5440 7520

5000 6000 7000 8000 5000 6000 7000 8000 full dataset reduced dataset

J¨ager,Yanovich et al. (T¨u) On phylogenetic dates MPI Jena 38 / 38