Even in Greenberg's scheme, there remain quite a number of individual families in South America, and they are much more interspersed than the more solid blocks in North America (especially Eastern North America). Hokan, Chibchan, Paezan, and Equatorial also occur in the North American map. |
This tree illustrates the claimed branches of the mega-family. Even many of the lower constituents, e.g. Hokan, are considered unproved by most linguists, as illustrated above. |
To motivate his groupings, Greenberg proposes cognate sets with quite a range of different meanings. These are in themselves not necessary implausible, but they can be verified only within a more rigorous consideration of systematic correspondences.
These numbers refer to the entry in the revised 2007 version of the Amerind dictionary. This next wide-ranging entry was in the original 1987 edition, although it's not in the new one under "drink":
It now occurs as a very long entry, # 852 "water/drink" */aq’ʷa ~ uqʼʷa/. No doubt this is related via Proto-World to Latin aqua. See the lists of random words. |
A striking example of changes in meaning within a well-understood family can be found in the history of English treacle, ultimately derived from PIE "wild animal". But these radical changes are established only because the history can be traced in a regular manner.
* Cognates that retain the original meaning are widespread in Balto-Slavic (e.g., Russian zv’er’ and Polish zwierzę "animal") and Lat. fera "wild beast" (cf. feral). ** Other examples of ellipsis of a head noun include English drive-in (theater) and single (base hit). |
The reason that we must look for systematic correspondences is that chance similarities are fairly easy to find. Here are some examples for English and Spanish: the languages are related, but these words are not.
Wikipedia has a long list of similar false cognates. |
On the other hand, true correspondences can be quite unusual thanks to a series of changes, for example English /f/ and Latin /k/.
By Grimm's Law, PIE */k/ > PGmc /x/. By a Middle English change, based on acoustic similarity, /x/ after a back vowel often became /f/. Latin /n/ infixation can also be seen in words such as frang-o "I break" vs. frag-ment-um and frac-t-us "broken". These words go back to PIE *bhreg- meaning "break", which has the same origin. |
Today we'll cover some examples of errors in the 1987 Amerind dictionary discussed in the readings. Some errors are resemblant forms that we know are derived from different roots: thus it's a mere coincidence, like day and día.
Others are not reconstructable within the family, and thus probably a later innovation/borrowing (esp. in Blackfoot, which is peripheral).
Some have incorrect or unmotivated meanings.
Examples from Ives Goddard (1987). Page numbers from Greenberg's book. Certain of these errors have been deleted in the 2007 version – including BELLY and ALL, as well as HAND in Kalapuya below – but it may be that new errors have been introduced. |
LARGE: ket’ul ‘to form large...’ (p.173)
HAND₁: -ketew (p.172)
OPEN₁: kæn (p.175)
Howard Berman (1992). |
ABOVE₂: am-efo ‘mountain’ (p.182, compared to aba, bax, etc.)
HAND: putukwi (p.151, with various p_t roots such as Chinook pote ‘arm’)
|