Perfect phylogenetic networks, and inferring language evolution

1. Perfect phylogenetic networks, and inferring language evolution Tandy Warnow The University of Texas at Austin (Joint work with Don Ringe, Steve Evans, and Luay Nakhleh)

2. Species phylogeny From the Tree of the Life Website,University of Arizona Orangutan Human Gorilla Chimpanzee

3. Possible Indo-European tree(Ringe, Warnow and Taylor 2000)

4. Controversies for Indo-European history • Subgrouping: Other than the 10 major subgroups, what is likely to be true? In particular, what about • Indo-Hittite • Italo-Celtic, • Greco-Armenian, • Anatolian + Tocharian, • Satem Core?

5. Historical Linguistic Data • A character is a function that maps a set of languages, L, to a set of states. • Three kinds of characters: • Phonological (sound changes) • Lexical (meanings based on a wordlist) • Morphological (especially inflectional)

6. Homoplasy-free evolution • When a character changes state, it changes to a new state not in the tree • In other words, there is no homoplasy (character reversal or parallel evolution) • First inferred for weird innovations in phonological characters and morphological characters in the 19th century. 0 0 1 0 0 0 0 1 1

7. Lexical characters can also evolve without homoplasy • For every cognate class, the nodes of the tree in that class should form a connected subset - as long as there is no undetected borrowing nor parallel semantic shift. • However, in practice, lexical characters are more likely to evolve homoplastically than complex phonological or morphological characters. 1 1 1 0 0 0 1 1 2

8. Differences between different characters • Lexical: most easily borrowed (most borrowings detectable), and homoplasy relatively frequent (we estimate about 25-30% overall for our wordlist, but a much smaller percentage for basic vocabulary). • Phonological: can still be borrowed but much less likely than lexical. Complex phonological characters are infrequently (if ever) homoplastic, although simple phonological characters very often homoplastic. • Morphological: least easily borrowed, least likely to be homoplastic.

9. Linguistic character evolution • Characters are lexical, phonological, and morphological. • Homoplasy is much less frequent than in biomolecular data: most changes result in a new state, and hence there is an unbounded number of possible states. • Borrowing between languages occurs, but can often be detected.

10. Our methods/models • Ringe & Warnow “Almost Perfect Phylogeny”: most characters evolve without homoplasy under a no-common-mechanism assumption (various publications since 1995) • Ringe, Warnow, & Nakhleh “Perfect Phylogenetic Network”: extends APP model to allow for borrowing, but assumes homoplasy-free evolution for all characters (Language, 2005) • Warnow, Evans, Ringe & Nakhleh “Extended Markov model”: parameterizes PPN and allows for homoplasy provided that homoplastic states can be identified from the data (to appear in Cambridge University Press) • Ongoing work: incorporating unidentified homoplasy.

11. First analysis: Almost Perfect Phylogeny • The original dataset contained 375 characters (336 lexical, 17 morphological, and 22 phonological). • We screened the dataset to eliminate characters likely to evolve homoplastically or by borrowing. • On this reduced dataset (259 lexical, 13 morphological, 22 phonological), we attempted to maximize the number of compatible characters while requiring that certain of the morphological and phonological characters be compatible. (Computational problem NP-hard.)

12. Indo-European Tree(95% of the characters compatible)

13. Initial analysis • Initial analysis of the IE dataset revealed that no perfect phylogeny for that dataset existed, even after careful screening. • Possible explanations: • Homoplasy • Polymorphism (e.g. rock/stone) • Mistakes in character coding • Borrowing (horizontal transmission)

14. Modelling borrowing: Networks and Trees within Networks

15. Perfect Phylogenetic Networks • An underlying tree + • additional contact edges • No cycles that involve • tree edges • Each character is • compatible on at least • one of the trees “inside” • the network 1 1 1 2 2 1 2 2 1

16. PPN Reconstruction Method Minimum Increment to a PPN (MIPPN): • Estimate the underlying “genetic” tree • Add a minimum number of contact edges to make all characters compatible (NP-hard to solve exactly even when the genetic tree is known, so we do exhaustive search on each candidate tree.)

17. The Indo-European (IE) Dataset • 24 languages • 294 characters: 22 phonological, 13 morphological, and 259 lexical • We examined five different “genetic” trees, one of which had a minimum number of incompatible characters (14 lexical characters)

18. Quality of Solutions Three mathematical criteria: • # characters incompatible with the “genetic” tree T • # additional contact edges needed to obtain a PPN from T • # borrowing events needed to make all characters compatible on the PPN Also: feasibility with respect to the archaeological and historical record

19. Our best PPN (Language, 2005)

20. Are we done? • We observed that of the three contact edges, only two are well-supported. If we eliminate that weakly supported edge, then we must explain the incompatibility of some characters (either through homoplasy or polymorphism). • Challenge: How to model polymorphism, homoplasy, borrowing, and genetic transmission?

21. Other work • Stochastic model of language evolution incorporating homoplasy, showing identifiability and efficient reconstructability (to appear Cambridge University Press) • Comparison of various methods on the IE dataset (to appear, Transactions of the Philological Society) • Modelling polymorphism (SIAM J. Computing, and ongoing) • Simulation study (ongoing)

22. For more information • Please see the Computational Phylogenetics for Historical Linguistics web site for papers, data, and additional material http://www.cs.rice.edu/~nakhleh/CPHL

23. Acknowledgements • The Program for Evolutionary Dynamics at Harvard • NSF, the David and Lucile Packard Foundation, the Radcliffe Institute for Advanced Studies, and the Institute for Cellular and Molecular Biology at UT-Austin. • Collaborators: Don Ringe, Steve Evans, and Luay Nakhleh.