First of all, Bantu and Niger-Congo don't have the same status, even approximately. Bantu is a bona fide family, while Niger-Congo is a loose cluster of a rather large number of families (including Bantu), established on the basis of typological similarities. No-one has yet demonstrated its genetic unity using generally accepted methods. _If_ it is real, then Bantu is just a subbranch of a subbranch of a subbranch in one of its branches. However, it's quite likely that the term "Niger-Congo" does not refer to a valid linguistic taxon at all.
 
You point to many /tVr-/ forms for "three" in (so-called) Niger-Congo laguages, but without analysing the comparative correspondences in a rigorous way you can't show that there was an old *tVr- word there, let alone deriving it from AN *telu. In fact, you quote rather selectively, choosing the forms that suit your suggestion but conveniently ignoring a number of /tat-/ forms in Niger-Congo. For example, in the Duru family (Cameroon, Nigeria) we have Sari tatu and Kotopo ta~to beside Sewe taare and Vere tariko.
 
As I have pointed out before, the frequent change of intervocalic *-t- into -d-, -l- or -r- in Bantu is a well-known fact; and many of the Bantu languages enjoy local or regional prestige, serving as means of wider communication. Even if Bantu and non-Bantu "Niger-Congo" languages are unrelated, the areal diffusion of Bantu numerals is easier to imagine than a borrowing from a more exotic source. The non-Bantu "NC" languages are spoken in _West_ Africa (Senegal, Guinea Bissau, Nigeria, Cameroon, etc.), very far away from the Indian Ocean.
 
As for AN presence in Africa, there is a pervasive Bantu substrate in Malagasy. The most natural explanation thereof is that proto-Malagasy speakers (AN traders arriving from southern Borneo) settled also on the eastern coast of Africa as they colonised the island, and that there was movement of people and diffusion of linguistic traits across the Mozambique Channel. The linguistic fingerprints of AN on the continent were eventually wiped out in the aftermath of the Bantu migration, but there was very likely some back-migration out of Africa, bringing Bantu-tinged AN dialects there.
 
The earliest archaeological traces of human settlement in Madagascar come from the fifth century. Also the mass extinction of the local megafauna began about that time. I find it hard to believe that an expert sea-people sailing to and fro between Africa and the Malay Archipelago (there is evidence of later contacts) should have missed the opportunity of colonising a virgin island, the fourth largest on this planet (Borneo being the third), that lay right on their course.
 
To sum up, the linguistic evidence, which suggests the fourth or fifth century as the likeliest date of the emergence of proto-Malagasy, squares well with the known history of Africa and Madagascar. Anyone who'd like to suggest substantial arrivals from the east at an earlier date would have problems explaining how on earth the AN seafolk managed to miss Madagascar. In other words, early Malagasy could be the only potential source of AN numerals (and other loans) in equatorial Africa: "isa, roa, telo, efatra, dimy, ...?". By the way, Proto-Bantu *tato '3' dates back to the fourth millennium BC or so.
 
Piotr
 
 
 
----- Original Message -----
From: tgpedersen@...
To: nostratic@yahoogroups.com
Sent: Friday, April 20, 2001 7:04 PM
Subject: [nostratic] Re: Afro-Austronesian numerals ?

> I don't question the possibility of AN migration(s) towards Africa.
We have Malagasy in Madagascar, of course, but that was a pretty
recent event (the position of Malagasy in the Malayo-Polynesian branch
is securely established and so is the approximate date of its
separation from its closest relatives).
>
> Piotr
>
True. But so is the Bantu expansion.
As I see it there are two possible cases.

1. Bantu is Niger-Congo
2. Bantu is not Niger-Congo

In either case, I can point to many *t-r- "3"'s in Niger-Congo
languages. It then doesn't become extremely important whether Bantu
"3" can be related to *t-r-.

In any case there is *t-r/l- in Austronesia, West Africa and IE (> AA
*T-l-T-). Now I would find that not very probable a priori, but in the
end it is of course also a matter of what other things you find
probable (eg. modes of transport).

Torsten