Does Basque Descend from Dogon?
I was recently asked to comment on media reports about the work of Basque linguist Jaime Martín Martín that purports to show that Basque descended from a West African language, Dogon. The Ethnologue classifies Dogon as 19 distinct yet closely related languages, all of which are spoken in central Mali, in the area shown in turquoise on the Ethnologue map reproduced on the left. There are some 503,000 speakers of Dogon, with the largest variety, Tomo Kan, spoken by 133,000 and the smallest, Ana Tinga, spoken by merely 500 people. The Dogon varieties are members of the Volta-Congo branch of the Atlantic-Congo sub-family of the Niger-Congo family, whose better-known members are the Bantu languages. If Martín is correct, it means that Basque is also a Niger-Congo language, a cousin of Swahili, Chichewa, and Xhosa. But as attempts to relate Basque to other languages are numerous yet virtually never successful (see for example the discussion of the alleged link between Basque and Georgian), it is worth looking at the sort of evidence that is presented to support Martín’s theory. It should be noted, however, that I have not been able to find Martín’s original paper, so unfortunately I have to rely on what has been reported by the popular media.
According to the reports, Martín compared both structural and lexical aspects of Basque and Dogon. From the structural point of view, it is claimed that the two languages are greatly alike, with the only difference being that Dogon “has no declensions or ergative subject”, which I can only interpret as saying that Dogon lacks case marking or ergative alignment. These are major differences, however, as they determine the overall structural “feel” of the language. The claim that Basque is “very similar” to Dogon is about as adequate as saying Basque is “very similar” to English or to Chinese. If anything, Dogon, lacking case marking or ergative alignment, is much more similar to Chinese—especially since both languages (or language families) are also tonal. Basque, on the other hand, is not tonal, which is another major difference between it and Dogon.
Another piece of evidence adduced in support of the Basque-Dogon link is that “three of the fourteen Dogon dialects showed exactly the same order of words in the sentence” as Basque. This is a piece of non-evidence if I ever saw one! Basque is a strict SOV language, but the SOV order is the most common cross-linguistically, more common than the SVO order, found in English. Nearly 45% of the world’s languages are SOV. Therefore, the chance that any randomly selected language has the same word order as Basque is nearly one in two—hardly a strong piece of evidence for relatedness.
It should also be noted that the order of major clausal constituents is not a strong type of evidence for language relatedness to begin with: related languages can have different word orders. For example, within the Indo-European language family we find SVO languages such as English, French, and Russian; SOV languages such as Farsi and Hindi (and other Indo-Iranian languages), and VSO languages such as Irish and Welsh (and other Celtic languages). If we look beyond the basic main clause word order, other differences between even more closely related languages become apparent. For example, in embedded clauses Irish retains the same VSO order as in its main clauses, while Welsh reverts to the SVO order. Similarly, German and Yiddish exhibit nearly identical order in main clauses, as both languages have the Verb-Second phenomenon. (German and Yiddish differ, however, in the OV/VO order, which only becomes apparent in the presence of an auxiliary, that is when the lexical verb is not subject to the Verb-Second constraint.) But in the embedded clauses, German and Yiddish differ drastically: in German, the finite (i.e. tense-bearing) verb comes last, while in Yiddish it comes second:
Er sagt daβ [die Kinder diesen Film gesehen haben]. [German]
he said that [the children this film seen have]
‘He said that the children have seen this film.’
Er hat gesagt az [di kinder haben gesen dem Film]. [Yiddish]
he has said that [the children have seen this film]
‘He said that the children have seen this film.’
Note that English, which is also closely related to German and Yiddish (all three are members of the West Germanic branch of the Germanic family), does not in general have the Verb-Second phenomenon, with some minor exceptions, discussed in my earlier post.
Another example of closely related languages differing in their word order patterns involves Russian (an East Slavic language) and Polish (a West Slavic language). While both languages allow verb-initial orders in main clauses, only in Polish can such verb-initial structures be embedded:
Zastanawiam się czy [pojedzie Mary do Hiszpanii tego lata]. [Polish]
I-wonder REFL whether goes Mary to Spain this summer
‘I wonder whether Mary is going to Spain this summer.’ [Santorini 1989: 173]
*V skazke govorit’sja čto [posadil ded repku]. [Russian]
in folktale says that planted grandpa.NOM turnip.ACC
intended: ‘The folktale says that the Grandpa planted a turnip.’
Numerous other examples of related languages differing in their word order patterns can be brought to bear on this issue, including the fact that most other Niger-Congo languages are SVO, rather than SOV, like Dogon. And when it comes to syntactic differences between closely related languages, a whole school of research emerged in recent years to deal with exactly those issues—micro-parametric syntax.
Similarly, macro-parameters, such as head/dependent marking (i.e. agreement vs. case marking) and ergative vs. nominative-accusative alignment, need not be a reliable argument in favor of language relatedness. For example, languages in the Caucasus all exhibit ergative alignment (or similar active alignment), but the three indigenous Caucasian language families differ as to whether they rely on case marking, agreement, or both. However, even languages within the same family can differ along those lines. For instance, English (like Dogon) lacks case marking, while Hindi and many other Indo-European languages (in particular, Indo-Iranian and Balto-Slavic languages) have case marking. Moreover, Hindi—much like Basque—exhibits ergative alignment, yet this trait is considerably rarer among Indo-European languages.
Finally, Martín’s claim of the relatedness between Dogon and Basque is said to be supported by lexical similarities. It is not clear, however, if these are true cognates or look-alikes. Therefore, this sort of evidence would not pass muster with any serious historical linguist—and perhaps not even as an answer to a Linguistics 101 quiz. Crucially, it is not clear that Martín has been able to establish any regular sound correspondences between Basque and Dogon. The two pairs of “similar” words presented in the media reports—bede / bide ‘way’ and beri / bero ‘hot’— do not allow to infer any such sound correspondences. If the first pair of words is supposed to illustrate the /e/-/i/ correspondence, why isn’t the second pair beri / biro?
All in all, I am not at all impressed by the evidence presented in support of the Basque-Dogon link. And crucially, my concern is not just quantitative, but qualitative: it is not that there is not enough evidence, but the “evidence” mentioned is of the wrong sort. There is nothing in these reports that would make me want to read the original paper or examine the evidence more closely.