Excerpt from change request document:
Kla-Dan is different enough from the Dan to make mutual intelligibility marginal; the percentage of cognates in Swadesh's 100-wordlist between Kla-Dan and different Dan dialects varies between 88 and 90%. Kla-Dan and Dan have no common literature; Kla- Dan speakers have an ethnolinguistic identity of their own, separate from the rest of Dan.
Dan and Kla-Dan can still be regarded as closely related languages, their speakers are aware of their common origin. Kla-Dan speakers do not accept their common identity with "Yakuba", the term commonly used in Côte d'Ivoire for the other Dan speakers, but they agree that they belong to the Dan community.
Further explanation from email:
About Dan and Kla-Dan: well... I see the point. For me, 90% of Swadesh's 100-word list is about the limit of inellegibility. In the Mande families, there are lots of varieties which are considered as separate languages although they have 95 to 98% of cognates in the 100- wordlist with other ones! If we take the 80% range as the limit, I'm afraid, many European languages will loose their status of languages...
It is true that when a 100-wordlis is collected badly, or based on unreliable and incomplete data, or do not take account of established regular correspondences, the rate is normally lower than in the contrary case. For example, in Ronald Long's dissertation (1971) "A comparative study of the Northern Mande languages" (which is still regarded by many as the main reference in the field), the rates are systematically lower than in my analysis (2009). Here are some examples:
Long 1971 Susu-Maninka 27-36%
etc. - this tendency is valid for every pair of languages without exception! Therefore, if I want to justify the necessity to split one language into two, I can simply take Long's data (instead of mine), and the critics will be convinced. Isn't it absurd?