Basing himself largely on areal and typological arguments, Güldemann (2010) claims that neither Proto-Niger-Congo nor Proto-Bantu had more than a "moderate" system of derivational verb suffixes ("extensions"), and that both proto-languages lacked inflectional verb prefixes. Although drawing largely on the same materials as Hyman (2004, 2007a, b), he arrives at the opposite conclusion that Niger-Congo languages which have such morphology, in particular Bantu and Atlantic, would have had to innovate multiple suffixation and prefixation. However, such hypotheses are weakened by two serious problems: (i) These proto-languages, which possibly reach back as far as 10,000–12,000 bp, have clearly had enough time for their morphosyntax to have cycled more than once. (ii) The areal properties of Güldemann's Macro-Sudan Belt most likely represent more recent innovations which have diffused after the Niger-Congo break-up. In this paper, I present further evidence that multiple suffixation and prefixation must have existed even in languages which have lost them. The general conclusion is that current areal distributions are largely irrelevant for long-range linguistic reconstruction.