|1. I thank
Emmanuel Drechsel, Penelope Drooker, Heather Hardy, John
Justeson, Geoff Kimball, Pat Kwachka, Jack Martin, Pamela Munro,
Dale Nicklas, Dean Snow, and the audiences at the Southern
Anthropological Society and University at Albany for their
comments and suggestions on this paper. All mistakes are my own.
The following abbrevations are used: : Al = Alabama, Ap =
Apalachee, Cr = Creek, Cs = Chickasaw, Ct = Choctaw, H =
Hitchiti, K = Koasati, MCt = Mississippi Choctaw, Mk = Mikasuki,
OS = Oklahoma Seminole, S = Seminole. The following orthographic
conventions are used in the citation of data from Muskogean
languages: nasalized vowels are indicated with a tilde; <ch> (in
Western Muskogean) and <c> (in other languages) represent […]; <lh>
represents ["] (a voiceless lateral fricative); and <sh> (in
Western Muskogean) represents [š].
The financial support of the Departments of Anthropology and
Linguistics and Cognitive Science at the University at Albany,
State University of New York is gratefully acknowledged.
2. In general this discussion
omits the extinct Muskogean languages Apalachee (Kimball 1987,
1988), Guale and Yamasee (Broadwell 1991), since the available
data are fragmentary. Haas (1949) argues that Apalachee is most
closely related to
[Page 42] Alabama and Koasati. Broadwell (1991) suggests that
Guale and Yamasee are most closely related to Creek.
3. Thus Kimball's (1989)
objection to Munro's classification on the grounds of the names
assigned to the branches is misguided.
4. Kimball is explicit about his
hypothesis of a ternary split in the text of his paper, but the
accompanying diagram shows all three possible branchings for
Central Muskogean with question marks (i.e. as a separate
branch, joined with Western Muskogean, and joined with Eastern
5. As Diamond (1992) notes,
critics of lexicostatistics are often nevertheless willing to
make estimates of time depth based on intuition.
6. In fact, there is some
possibility that pãshi/pãshi' may also be cognate with the other
words for `hair'. The etymology of these forms is something of a
mystery, but one derivation might be PM *ik w a `head' + *hisi
`hair'. While PM *k w generally develops into Western Muskogean
/b/, there are some cases in Choctaw where initial /p/ and
/b/alternate, e.g. pichilli/bichilli `to ooze out', bishlichi/wishlichi
`to milk' and pishi `to suck'.
7. Presumably this is due to the
environment where Mikasuki is currently spoken.
8. Muskogean seems comparable in
diversity with the [Page 43] Romance family, which has a time
depth of about 2500 years. This may suggest that the date for
Proto-Muskogean is somewhat more recent than 960 BC figure
given. A date of ca. 600 BC is within the range of error of the
lexicostatistical calculations and consistent with the
comparison to Romance.
9. Nicklas (this volume)
suggests that the fact that the figures for an Alabama-Creek
split and a Choctaw-Creek split differ from each other shows
that lexicostatistical methods are invalid. This is a mistaken
interpretation of the data.
The figures in the table above attempt to give us some
quantitative measure of the degree of similarity between the
Muskogean languages. That similarity is then used to estimate
the degree of separation between the languages. However, some
portion of that similarity is due to a distant common history,
while another portion of the similarity is due to more recent
influence and borrowing. It is entirely consistent with the
results above to claim that the Choctaw and Alabama languages
are equally distant from Creek, but that Alabama shows a greater
modern similarity due to more recent Creek influence. In
deciding which of the estimates of the separation date is more
likely to be correct, we should thus prefer the more distant
figure suggested by the Choctaw-Creek comparison. [Page 44]
10. The list below includes all
the terms for flora and fauna listed in Munro et al. (1991),
with the following exceptions: I did not include BIRD, BIRD SP.,
FISH SP, FISH (because they are not specific enough to give us
useful information) or obvious borrowings from European
languages (BACON, CAT, COFFEE, COW, GOAT, OKRA, RICE, TOMATO,
I reject the proposed cognate sets for BEAN, BEAVER, HUCKLEBERRY
2, PEANUT as improbable.
The cognate set for CHICKEN is a special problem. There is a
similar word for `chicken' in many languages, but it must have
originally applied to some other sort of bird, since the chicken
is a European introduction.
Some sets contain lexical material duplicated in other sets, and
I have attempted to include only one set in such cases. I
include OPOSSUM, but not HOG, since they are from the same root.
I exclude OWL 4, since it is the same root as HOOT OWL. I do not
include the compounds ROADRUNNER (`fast bird') or WHALE (`water
11. The following discussion
relies heavily on the discussion of the cognate sets in Munro et
12. As Martin (1987:117) notes,
the phonology of the Mk. nokosi `bear' is unusual for a Mikasuki
word, since we would expect lengthening of the initial syllable
in words of this shape. This strengthens the case for treating
this word as [Page 45] a loan.
13. In many of the other
languages, `watermelon' appears to be a compound based on
`pumpkin' + a root /tal(ak)/ which may mean `lie down'.
14. Al, K cici probably
originates as a children's word for penis. The origin of Cs.
inkilish is obscure, and it is not clear that it is cognate to
the other items.
15. Data for this appendix were
provided by Heather Hardy (Alabama), Jack Martin (Creek and
Mikasuki), and Pamela Munro (Chickasaw). Choctaw data comes from
Byington (1915) and Broadwell (1987b). [Page 46]