The genealogical unity and internal classification of the Niger-Congo phylum remain highly doubtful. The lack of progress in this domain is due to several factors, including the very high amount of languages involved, a lack of data on languages spoken in some crucial areas (most notably Nigeria) and the existence of large linguistic spread zones. Using the lexical database RefLex and its tools, Guillaume Segerer and Konstantin Pozdniakov have recently developed an innovative methodology for hypothesis creation based on the detection of recurring patterns of homophony. This methodology opens exciting new avenues for historical linguistics on the African continent.

The discovery of regular homophonies, defined as sets of lexemes that have the same multiple unrelated meanings, provides a very strong research hypothesis for the establishment or confirmation of genealogical relatedness between languages, as well as for the search of regular phonological correspondences. This can be illustrated with a lexeme that has the same pattern of homophony in Kare [kbn] (CAR, classified as “Adamawa”), Khana [ogo] (Nigeria, Delta-Cross) and Kete [kcv] (DRC, Bantu), as shown in Table 1.

Being statistically highly unlikely, such series give a very strong phylogenetic signal. Consequently, the sounds involved illustrate phonetic correspondences that can be qualified as regular, even if only a single example of homophony were found. These correspondences can then be tested with other lexemes in the same languages, as briefly illustrated in Table 2.

Only the series ‘ear’ and ‘ashes’ are complete for the three languages, but these series allow us to confirm the validity of the incomplete series. 

The undergraduate researcher in this project will use the RefLex database to systematically search for patterns of regular homophony. He or she will then pursue the ensuing hypotheses in terms of classification and reconstruction, in collaboration with the specialists of Niger-Congo subgroups who are working at LLACAN.

The successful candidate will contribute to the delimitation and internal classification of the Niger-Congo phylum and to the reconstruction of some of its sub-branches by pursuing the method of regular homophony detection in the large lexical database RefLex.

Research context: LABEX-EFL, Strand 3, research project GL7 Reconstruction, genealogy, typology and grammatical description in the world’s two biggest phylums: Niger-Congo and Austronesian (dir. Isabelle Bril, Alexandre François, Mark van de Velde)

Institutional context: INALCO & LLACAN. The LLACAN is a vibrant community of experts in African languages and linguistics, located in the southern suburbs of Paris (reachable by metro and other public transport). The candidate will be enrolled as a PhD student at Inalco and will be supervised by Guillaume Segerer, Mark Van de Velde and/or Konstantin Pozdniakov. 


  • An MA degree in Linguistics before starting the contract;
  • An interest in historical linguistics, African languages and/or linguistic databases;
  • Working knowledge of English and French.

Contact: Mark Van de Velde (