|Central Africa, north-central Africa and East Africa|
|Linguistic classification||If valid, one of the world's primary language families|
|ISO 639-2 / 5||ssa|
Map showing the distribution of Nilo-Saharan languages
The Nilo-Saharan languages are a proposed family of African languages spoken by some 50-60 million people, mainly in the upper parts of the Chari and Nile rivers, including historic Nubia, north of where the two tributaries of the Nile meet. The languages extend through 17 nations in the northern half of Africa: from Algeria to Benin in the west; from Libya to the Democratic Republic of the Congo in the centre; and from Egypt to Tanzania in the east.
As indicated by its hyphenated name, Nilo-Saharan is a family of the African interior, including the greater Nile Basin and the Central Sahara Desert. Eight of its proposed constituent divisions (excluding Kunama, Kuliak, and Songhay) are found in the modern two nations of Sudan and South Sudan, through which the Nile River flows.
In his book The Languages of Africa (1963), Joseph Greenberg named the group and argued it was a genetic family. It contains the languages which are not included in the Niger-Congo, Afroasiatic or Khoisan groups. Although some linguists have seen the phylum as "Greenberg's wastebasket", into which he placed all the otherwise unaffiliated non-click languages of Africa, specialists in the field have accepted its reality since Greenberg's classification.  Its supporters accept that it is a challenging proposal to demonstrate but contend that it looks more promising the more work is done.
Some of the constituent groups of Nilo-Saharan are estimated to predate the African neolithic. Thus, the unity of Eastern Sudanic is estimated to date to at least the 5th millennium BC. Nilo-Saharan genetic unity would necessarily be much older still and date to the late Upper Paleolithic.
This larger classification system is not accepted by all linguists, however. Glottolog (2013), for example, a publication of the Max Planck Institute in Germany, does not recognise the unity of the Nilo-Saharan family or even of the Eastern Sudanic branch; Georgiy Starostin (2016) likewise does not accept a relationship between the branches of Nilo-Saharan, though he leaves open the possibility that some of them may prove to be related to each other once the necessary reconstructive work is done.
The constituent families of Nilo-Saharan are quite diverse. One characteristic feature is a tripartite singulative-collective-plurative number system, which Blench (2010) believes is a result of a noun-classifier system in the protolanguage. The distribution of the families may reflect ancient water courses in a green Sahara during the Neolithic Subpluvial, when the desert was more habitable than it is today.
Within the Nilo-Saharan languages are a number of languages with at least a million speakers (most data from SIL's Ethnologue 16 (2009)). In descending order:
Some other important Nilo-Saharan languages under 1 million speakers:
The total for all speakers of Nilo-Saharan languages according to Ethnologue 16 is 38-39 million people. However, the data spans a range from ca. 1980 to 2005, with a weighted median at ca. 1990. Given population growth rates, the figure in 2010 might be half again higher, or about 60 million.
The Saharan family (which includes Kanuri, Kanembu, the Tebu languages, and Zaghawa) was recognized by Heinrich Barth in 1853, the Nilotic languages by Karl Richard Lepsius in 1880, the various constituent branches of Central Sudanic (but not the connection between them) by Friedrich Müller in 1889, and the Maban family by Maurice Gaudefroy-Demombynes in 1907. The first inklings of a wider family came in 1912, when Diedrich Westermann included three of the (still independent) Central Sudanic families within Nilotic in a proposal he called Niloto-Sudanic; this expanded Nilotic was in turn linked to Nubian, Kunama, and possibly Berta, essentially Greenberg's Macro-Sudanic (Chari-Nile) proposal of 1954.
In 1920 G. W. Murray fleshed out the Eastern Sudanic languages when he grouped Nilotic, Nubian, Nera, Gaam, and Kunama. Carlo Conti Rossini made similar proposals in 1926, and in 1935 Westermann added Murle. In 1940 A. N. Tucker published evidence linking five of the six branches of Central Sudanic alongside his more explicit proposal for East Sudanic. In 1950 Greenberg retained Eastern Sudanic and Central Sudanic as separate families, but accepted Westermann's conclusions of four decades earlier in 1954 when he linked them together as Macro-Sudanic (later Chari-Nile, from the Chari and Nile Watersheds).
Greenberg's later contribution came in 1963, when he tied Chari-Nile to Songhai, Saharan, Maban, Fur, and Koman-Gumuz and coined the current name Nilo-Saharan for the resulting family. Lionel Bender noted that Chari-Nile was a historical artifact of the discovery of the family and did not reflect an exclusive relationship between these languages, and the group has been abandoned, with its constituents becoming primary branches of Nilo-Saharan--or, equivalently, Chari-Nile and Nilo-Saharan have merged, with the name Nilo-Saharan retained. When it was realized that the Kadu languages were not Niger-Congo, they were commonly assumed to therefore be Nilo-Saharan, but this remains somewhat controversial.
Progress has been made since Greenberg established the plausibility of the family. Koman and Gumuz remain poorly attested and are difficult to work with, while arguments continue over the inclusion of Songhai. Blench (2010) believes that the distribution of Nilo-Saharan reflects the waterways of the wet Sahara 12,000 years ago, and that the protolanguage had noun classifiers, which today are reflected in a diverse range of prefixes, suffixes, and number marking.
Dimmendaal (2008) notes that Greenberg (1963) based his conclusion on strong evidence and that the proposal as a whole has become more convincing in the decades since. Mikkola (1999) reviewed Greenberg's evidence and found it convincing. Roger Blench notes morphological similarities in all putative branches, which leads him to believe that the family is likely to be valid.
Koman and Gumuz are poorly known and have been difficult to evaluate until recently.[vague] Songhay is markedly divergent, in part due to massive influence from the Mande languages . Also problematic are the Kuliak languages, which are spoken by hunter-gatherers and appear to retain a non-Nilo-Saharan core; Blench believes they may have been similar to Hadza or Dahalo and shifted incompletely to Nilo-Saharan.
Anbessa Tefera and Peter Unseth consider the poorly attested Shabo language to be Nilo-Saharan, though unclassified within the family due to lack of data; Dimmendaal and Blench consider it to be a language isolate on current evidence. Proposals have sometimes been made to add Mande (usually included in Niger-Congo), largely due to its many noteworthy similarities with Songhay rather than with Nilo-Saharan as a whole, however this relationship is more likely due to a close relationship between Songhay and Mande many thousands of years ago in the early days of Nilo-Saharan, so the relationship is probably more one of ancient contact than a genetic link .
The extinct Meroitic language of ancient Kush has been accepted by linguists such as Rille, Dimmendaal, and Blench as Nilo-Saharan, though others argue for an Afroasiatic affiliation. It is poorly attested.
There is little doubt that the constituent families of Nilo-Saharan--of which only Eastern Sudanic and Central Sudanic show much internal diversity--are valid groups. However, there have been several conflicting classifications in grouping them together. Each of the proposed higher-order groups has been rejected by other researchers: Greenberg's Chari-Nile by Bender and Blench, and Bender's Core Nilo-Saharan by Dimmendaal and Blench. What remains are eight (Dimmendaal) to twelve (Bender) constituent families of no consensus arrangement.
Gumuz was not recognized as distinct from neighboring Koman; it was separated out (forming "Komuz") by Bender (1989).
Lionel Bender came up with a classification which expanded upon and revised that of Greenberg. He considered Fur and Maban to constitute a Fur-Maban branch, added Kadu to Nilo-Saharan, removed Kuliak from Eastern Sudanic, removed Gumuz from Koman (but left it as a sister node), and chose to posit Kunama as an independent branch of the family. By 1991 he had added more detail to the tree, dividing Chari-Nile into nested clades, including a Core group in which Berta was considered divergent, and coordinating Fur-Maban as a sister clade to Chari-Nile.
Bender revised his model of Nilo-Saharan again in 1996, at which point he split Koman and Gumuz into completely separate branches of Core Nilo-Saharan.
Christopher Ehret came up with a novel classification of Nilo-Saharan as a preliminary part of his then-ongoing research into the macrofamily. His evidence for the classification was not fully published until much later (see Ehret 2001 below), and so it did not attain the same level of acclaim as competing proposals, namely those of Bender and Blench.
By 2000 Bender had entirely abandoned the Chari-Nile and Komuz branches. He also added Kunama back to the "Satellite-Core" group and simplified the subdivisions therein. He retracted the inclusion of Shabo, stating that it could not yet be adequately classified but might prove to be Nilo-Saharan once sufficient research has been done. This tentative and somewhat conservative classification held as a sort of standard for the next decade.
Ehret's updated classification was published in his book A Historical-Comparative Reconstruction of Nilo-Saharan (2001). This model is notable in that it consists of two primary branches: Gumuz-Koman, and a Sudanic group containing the rest of the families (see Sudanic languages § Nilo-Saharan for more detail). Also, unusually, Songhay is well-nested within a core group and coordinate with Maban in a "Western Sahelian" clade, and Kadu is not included in Nilo-Saharan. Note that "Koman" in this classification is equivalent to Komuz, i.e. a family with Gumuz and Koman as primary branches, and Ehret renames the traditional Koman group as "Western Koman".
With a better understanding of Nilo-Saharan classifiers, and the affixes or number marking they have developed into in various branches, Blench believes that all of the families postulated as Nilo-Saharan belong together. He proposes the following tentative internal classification, with Songhai closest to Saharan, a relationship that had not previously been suggested:
By 2015, and again in 2017, Blench had refined the subclassification of this model, linking Maban with Fur, Kadu with Eastern Sudanic, and Kuliak with the node that contained them, for the following structure:
Gerrit J. Dimmendaal suggests the following subclassification of Nilo-Saharan:
The large Northeastern division is based on several typological markers:
In summarizing the literature to date, Hammarström et al. in Glottolog do not accept that the following families are demonstrably related with current research:
Georgiy Starostin (2016), using lexicostatistics based on Swadesh lists, is more inclusive than Glottolog, and in addition finds probable and possible links between the families that will require reconstruction of the protolanguages for confirmation.
In addition to the families listed in Glottolog (previous section), Starostin considers the following to be established:
A relationship of Nyima with Nubian, Nara, and Tama (NNT) is considered "highly likely" and close enough that proper comparative work should be able to demonstrate the connection if it's valid, though it would fall outside NNT proper (see Eastern Sudanic languages).
Other units that are "highly likely" to eventually prove to be valid families are:
In summary, at this level of certainty, "Nilo-Saharan" constitutes ten distinct and separate language families: Eastern Sudanic, Central Sudanic - Kadu, Maba-Kunama, Komuz, Saharan, Songhai, Kuliak, Fur, Berta, and Shabo.
Possible further "deep" connections, which cannot be evaluated until the proper comparative work on the constituent branches has been completed, are:
There are faint suggestions that Eastern and Central Sudanic may be related (essentially the old Chari–Nile clade), though that possibility is "unexplorable under current conditions" and could be complicated if Niger–Congo were added to the comparison. Starostin finds no evidence that the Komuz, Kuliak, Saharan, Songhai, or Shabo languages are related to any of the other Nilo-Saharan languages. Mimi-D and Meroitic were not considered, though Starostin had previously proposed that Mimi-D was also an isolate despite its slight similarity to Central Sudanic.
In a follow up study published in 2017, Starostin reiterated his previous points as well as explicitly accepting a genetic relationship between Macro-East Sudanic and Macro-Central Sudanic. Starostin names this proposal "Macro-Sudanic" 
Proposals for the external relationships of Nilo-Saharan typically center on Niger-Congo: Gregersen (1972) grouped the two together as Kongo-Saharan. However, Blench (2011) proposed that the similarities between Niger-Congo and Nilo-Saharan (specifically Atlantic-Congo and Central Sudanic) are due to contact, with the noun-class system of Niger-Congo developed from, or elaborated on the model of, the noun classifiers of Central Sudanic.
Nilo-Saharan languages present great differences, being a highly diversified group. It has proven difficult to reconstruct many aspects of Proto-Nilo-Saharan. Two very different reconstructions of the proto-language have been proposed by Lionel Bender and Christopher Ehret.
The consonant system reconstructed by Bender for Proto-Nilo-Saharan is:
|plosive||voiceless||*t, *t2||*k, *k?|
The phonemes /*d2, *t2/ correspond to coronal plosives, the phonetic details are difficult to specify, but clearly, they remain distinct from /*d, *t/ and supported by many phonetic correspondences (another author, C. Ehret, reconstructs for the coronal area the sound [d?], [?] and [t?], [?] which perhaps are closer to the phonetic detail of /*d2, *t2/, see infra)
Bender gave a list of about 350 cognates and discussed in deep the grouping and the phonological system proposed by Ch. Ehret. Blench (2000) compares both systems (Bender's and Ehret's) and prefers the former because it is more secure and is based in more reliable data. For example, Bender points out that there is a set of phonemes including implosives /*?, *?, *?, *?/, ejectives /*p', *t', (*s'), *c', *k'/ and prenasal constants /*mb, *nd, (*nt), *ñ?, *?g/, but it seems that they can be reconstructed only for core groups (E, I, J, L) and the collateral group (C, D, F, G, H), but not for Proto-Nilo-Saharan.
Christopher Ehret used a less clear methodology and proposed a maximalist phonemic system:
Ehret's maximalist system has been criticized by Bender and Blench. These authors state that the correspondences used by Ehret are not very clear and because of this many of the sounds in the table may only be allophonic variations.
Dimmendaal (2016) cites the following morphological elements as stable across Nilo-Saharan: