The core is held together by regular sound correspondences in lexical items between subfamily languages, less systematic verbal extensions and noun class systems Stewart, John M. 2002 , Stewart, John M. 2005 , Becher, Jutta 2002 , Mukarovsky, Hans G. 1976-1977 , Pozdniakov, Konstantin Igorevich and Segerer, Guillaume 2004 , Hyman, Larry M. 2007 , Holst, Jan Henrik 2007 Tom Güldemann 2018: 154-174 , 177-179, 180-189, 194-199, 200-213, 213-223, 223-231. Mande Tom Güldemann 2018: 189-192 , Kordofanian, Ijoid Tom Güldemann 2018: 174-177 and Dogon (cf. Tom Güldemann 2018: 192-193 ) have not yet been shown to stand up to these criteria wherefore they are excluded. Regarding Siamou, as Prost Prost, André 1964: 345 noted, apart from obvious loanwords, Siamou is neither Mande nor Gur. Contra Person, Yvonne 1966 , although there are typological matches, Simaou differs from Kru throughout the comparison, and Siamou can thus not be verified as a Kru language Zogbo, Lynell Marchese 2012 Tom Güldemann 2018: 177-180 . It lacks any of the diagnostic features of the Atlantic-Congo family and must thus be regarded as unrelated until such evidence is brought forward. The Pre or Mbre language in Côte d'Ivoire, known from wordlists only, is Atlantic-Congo but cannot be assigned to a specific branch at this time Blench, Roger 2006 , Boukari, Oumarou 2009 Tom Güldemann 2018: 180 . The Mpra (= Mpre) language in Ghana has cognates with Atlantic-Congo especially Guang Goody, Jack R. 1963 , including numerals 2-5, but the bulk of the little vocabulary that is known, is not Atlantic-Congo Blench, Roger 2012 . On the grounds that the numerals are less likely to be borrowed, and that lexical innovation may produce vocabulary that looks unrelated to anything else, we count Mpra as an Atlantic-Congo language.