Abstract: That the Mande family belongs to the Niger-Congo macro-family was first proposed by Joseph Greenberg in the middle of the 20th century. Over the last two decades this affiliation has regularly been questioned, the main argument for skepticism being the well-known fact that there is no trace of any noun class in Mande. Our study proves that these doubts are groundless as far as the lexicon is concerned. Our conclusion is based on the following arguments, which appear to be sufficient for the establishment of a relationship between Mande and Bantu and, consequently, for the inclusion of Mande in Niger-Congo:

• In the basic vocabulary, the average cognacy rate of contemporary Mande languages with Proto-Bantu is higher than the rate for other families whose membership in the Niger-Congo macro-family is uncontested.

• In the basic vocabulary, the number of phonetic similarities actually observed far exceeds the number expected on the basis of random coincidence alone.

• In the basic vocabulary, Proto-Bantu and Proto-Mande are considerably closer to each other than contemporary Bantu and Mande languages are.

• Phonetic correspondences between Proto-Bantu and Proto-Mande are not only regular, but also proportional, which excludes the possibility that they are merely fortuitous.

• In the basic vocabulary, the distribution of roots by their stability in the Mande family displays a strong positive correlation with their distribution in the other Niger-Congo families.

If we establish cognacy between specific Mande and Bantu roots, these roots can be regarded as candidates for membership in the Niger-Congo common lexical stock. We thus provide materials for a Proto-Niger-Congo comparative etymological dictionary: 207 common roots and their probable reflexes in contemporary languages and lower-level proto-languages are proposed for future discussion by specialists.