Indo-European Lexical Cognacy Database (IELex) and ‘Good Enough’ Tree

This repository contains data and scripts for producing a Bayesian phylogenetic tree sample of for the Indo-European family which is good enough for use in phylogenetic comparative methods. It should be considered as the current version of the "Indo-European Lexical Cognacy Database" from Dunn et al. (2011), popularly known as "IELex". The "Indo-European Cognate Relationships" (IE-CoR) database project, now based at the Max Planck Institute for Evolutionary Anthropology, has been working on a much improved database originally based on this data, but has not had any public data release so far. In the meantime, this is the best we can offer, as an alternative to the original website (http://ielex.mpi.nl) which is now offline.

Design principles

• The cognate coding is borrowed from Andrew Garrett's corrections of the IELex database (published in supplementary materials to Chang et al. 2015).
• Generated with standard BEAST2, without further customisation (Bouckaert et al., 2019). This means we cannot use the Chang et al 2015 notion of ancestry constraints (and there are theoretical and empirical reasons that make these controversial in any case).
• Ultimately we will offer control files generated using the beastling package (Maurits et al. 2017), which would make it very easy for users to change data and model settings; unfortunately we have found some bugs in the current iteration of beastling which makes it unsuitable.

The "Broad data set"

• Use the Bouckaert et al. (2012) language sample minus sparsely attested languages, following Chang et al. (2015). This is a tradeoff between quality of the wordlist and quantity of languages in the sample.

• Slightly truncated meaning list:

  The broad data set consists of ninety-four languages and 197 meaning classes. Of the 207 meaning classes in IELEX, three (‘blow’, ‘father’, ‘mother’) were excluded as being especially susceptible to sound symbolism, and seven others were excluded because they were unattested in more than 30% of the languages, probably due to the fact that they are not in the Dyen data set, which was constructed around the 200 items proposed by Swadesh (1952). (Chang et al 2015:213)

  Dropped meanings are: "blow", "breast", "father", "fingernail", "full", "horn", "knee", "moon", "mother", "round".

• Please note that the broad data file only includes information on cognancy, without the form or any phonological transcription of the lexemes.

Calibrations

Calibrations are derived from Chang et al. (2015) for individual languages (but 0-50 year calibrations on the modern languages are not included) and Bouckaert et al. (2012) for the family level calibrations. These are listed in beastling-compatible format ("X - Y" indicates a uniform distribution between X and Y; for normal and rlognormal the first parameter is the mean and second parameter is standard deviation).

Language/Clade	Calibration	Source
Ancient Greek	2400-2500	Chang et al. (2015)
Avestan	2450-2550	Chang et al. (2015)
Balto Slavic	normal(3100.0, 600.0)1	Bouckaert et al. (2012)
Celtic	1200 + rlognormal(2000.0, 0.6)	Bouckaert et al. (2012)
Classical Armenian	1500-1600	Chang et al. (2015)
Cornish	200-400	Chang et al. (2015)
French Iberian	normal(1400.0, 100.0)	Bouckaert et al. (2012)
Gothic	1625-1675	Chang et al. (2015)
Greek split	> 2500.0	Bouckaert et al. (2012)
Hittite	3300-3500	Chang et al. (2015)
Indic	2150 + rlognormal(1000.0, 1.0)	Bouckaert et al. (2012)
Indo-Iranian	3000-10000	Bouckaert et al. (2012)
Iranian	2600 + rlognormal(400.0, 0.8)	Bouckaert et al. (2012)
Latin	2100-2200	Chang et al. (2015)
Latin Romance	normal(2000.0, 135.0)	Bouckaert et al. (2012)
Latvian, Lithuanian	normal(1350.0, 25.0)	Bouckaert et al. (2012)
Northwest Germanic	normal(1875.0, 67.0)	Bouckaert et al. (2012)
Old Church Slavic	950-1050	Chang et al. (2015)
Old English	900-1050	Chang et al. (2015)
Old High German	1100-1200	Chang et al. (2015)
Old Irish	1100-1300	Chang et al. (2015)
Old Prussian	500-600	Chang et al. (2015)
Old West Norse	750-850	Chang et al. (2015)
Slavic	1200 + rlognormal(300.0, 0.6)	Bouckaert et al. (2012)
Sogdian	1200-1400	Chang et al. (2015)
Tocharian A	1200-1500	Chang et al. (2015)
Tocharian B	1200-1500	Chang et al. (2015)
Tocharic	1650 + rlognormal(200.0, 0.9)	Bouckaert et al. (2012)
Vedic Sanskrit	3000-3500	Chang et al. (2015)
West Germanic	normal(1550.0, 25.0)	Bouckaert et al. (2012)
originate(Brythonic)	normal(1500.0, 25.0)	Bouckaert et al. (2012)

Tree samples

IE-trees-v1

The control file IE-trees-v1/ie-v1.xml was made manually by Tiago Tresoldi, following as close as possible the model generated by beastling. The model was built in BEAUTi, following the parameters of the original beastling configuration. We use a Binary Covarion model for a coalescent tree with constant population. Missing data is encoded as ?; we only used concepts that had at least 90% coverage across the doculects. Concepts were divided in four different partitions, each with its individual site model, with each partition holding roughly the same number of concepts each and concept divided by the cardinality of their corresponding sets of individual cognate sets (25%, 50%, 75%, and 100% percentiles). We used a chain length of 10^8 and a burn-in of 50%. The tree includes multimonophyletic constraints, as given in the corresponding Newick tree, and date calibrations for both tips and splits as given above.

Files

The data/ directory holds the files most researchers will be interested in:

• ielex.csv is a single table in long format containing the essential database information, given one entry per row, each with the associated language, concept, and cognate set. It also carries the glottocode (Hammarström et al., 2021) corresponding to each language; concept glosses are given via the corresponding Concepticon (List et al., 2021) cognate set. It is the file most people will want when searching for "IELex data".

• concepts.csv is a list mapping all concepts used in IELex to their corresponding concept set in Concepticon (List et al., 2021), giving both the gloss and the id.

• ielex.nex is a NEXUS file with the information from the table above, with ascertainment correction columns, charstate labels, and assumptions. The file is built with build/build_nexus.py.

• ielex.nn.pdf and ielex.nn.png are a PDF and a high-resolution PNG from the NEXUS file above, generated with SplitsTree (Huson & Bryant, 2006).

• ielex.mcc.tre is the Maximum Clade Credibility ("consensus") tree from our latest analysis (currently IE-tree-v1). The first 50% of the sample was removed as burn-in.

• ielex.mcc.pdf is a graphic rendering of the tree above, generated with FigTree.

• shared_cognates.tsv is a table with percentage of shared cognates per doculect pair, as frequently requested by other researchers.

The build/ directory holds the files used for preparing the phylogenetic reconstruction.

The IE-trees-v1/ directory holds the files related to the phylogenetic reconstruction, including model, logs, and state. As stated above, a summary of the main output, as ielex.mcc.tre, is found in the data/ directory. For statistical purposes researchers are likely to want the entire (unsummarised) tree sample, IE-trees-v1/ie-v1.nex.

Tree and network

Authors and citation

The data and analysis was prepared within the context of the Cultural Evolution of Texts project, with funding from the Riksbankens Jubileumsfond (grant agreement ID: MXM19-1087:1).

If you use the data from this repository, please cite it as:

Michael Dunn & Tiago Tresoldi. (2022). evotext/ielex-data-and-tree: IELex data and tree (Version r20220328) [Data set]. Zenodo. https://doi.org/10.5281/zenodo.5556800

In BibTeX:

@dataset{Dunn_5556800,
  author       = {Michael Dunn and Tiago Tresoldi},
  title        = {{evotext/ielex-data-and-tree: IELex data and tree (2022/03/28)}},
  month        = mar,
  year         = 2022,
  publisher    = {Zenodo},
  version      = {r20211108},
  doi          = {10.5281/zenodo.5556800},
  url          = {https://doi.org/10.5281/zenodo.5556800}
}

References

Bouckaert R. R., Lemey P., Dunn M., Greenhill S. J., Alekseyenko A. V., Drummond A. J., Gray R. D., Suchard M. A. & Atkinson Q. D. 2012. Mapping the Origins and Expansion of the Indo-European Language Family. Science, 337(6097), 957-960.

Bouckaert R., Vaughan T. G., Barido-Sottani J., Duchêne S., Fourment M., Gavryushkina A., et al. (2019) BEAST 2.5: An advanced software platform for Bayesian evolutionary analysis. PLoS computational biology, 15(4), e1006650.

Chang W., Cathcart C., Hall D. & Garrett A. 2015. Ancestry-constrained phylogenetic analysis supports the Indo-European steppe hypothesis. Language, 91(1):194-244.

Dunn M., Ludewig J. et al. 2011. IELex, the Indo-European Lexical Cognacy Database. Max Planck Institute for Psycholinguistics: Nijmegen.

Hammarström H., Forkel R., Haspelmath M. & Bank S. 2021. Glottolog 4.4. Leipzig: Max Planck Institute for Evolutionary Anthropology. https://doi.org/10.5281/zenodo.4761960

Huson D. H. & Bryant D. 2006. Application of Phylogenetic Networks in Evolutionary Studies, Mol. Biol. Evol., 23(2):254-267.

List, J.-M., Rzymski C., Greenhill S. J., Schweikhard N., Pianykh K., Tjuka A., Hundt C. & Forkel R. (eds.). 2021. CLLD Concepticon 2.5.0 [Data set]. Zenodo. https://doi.org/10.5281/zenodo.4911605

Maurits L., Forkel R., Kaiping G. A. & Atkinson Q. D. 2017. BEASTling: A Software Tool for Linguistic Phylogenetics Using BEAST 2. PLOS ONE 12 (8): e0180908. https://doi.org/10.1371/journal.pone.0180908.

Rambaut A. Figtree v.1.4.4. http://tree.bio.ed.ac.uk/software/figtree/

Footnotes

1. Note that, as per source, Balto_Slavic distribution should be truncated from 2000-3400. This is not done in the model we are presenting here. ↩