_one_ atamek attakon kemmis.
_two_ malgok alluk nittanu.
_three_ pinajut kankun tshushquat.
_four_ istamat thitshin tshashcha.
_five_ tatlimat sshang koomdas.
No wonder that the tongues thus represented seem unlike.
But let us go farther--in the first place remembering that, in most
cases, it is only as far as _five_ that the ruder languages have
distinct numerals; in other words that from _six_ onwards they count
upon the same principle as _we_ do after ten, _i. e._ they join
together some two, or more, of the previous numerals; even as we, by
adding _seven_ and _ten_, make _seven-teen_. The exact details, of
course, differ; the general principle, however, is the same viz.: that
after _five_ the numerals become, more or less, compound, just as, with
us, they become so after _ten_.
With this preliminary observation let us ask what will be the
Kamskadale for _seven_ when _nittanu_ = _two_, and _kumdas_ = _five_.
The answer is either _nittanu-kumdas_ or _kumdas-nittanu_. But
the Kamskadale happens to have a separate word for _six_, viz.
_kiekoas_. What then? The word for _seven_ may be one of two things:
it may be either = 6 + 1, or 5 + 2. The former being the case,
and _kemmis_ = _one_, the Kamskadale for _seven_ should be either
_kemmis-kilkoas_ or _kilkoas-kemmis_. But it is neither one nor the
other. It is _ittakh-tenu_. Now as _eight_ = _tshok-tenu_ we know this
word to be compound. But what are its elements? We fail to find
them amongst the simpler words expressive of _one_, _two_, _three_,
_four_, _five_. We fail to find them amongst these if we look to the
Kamskadale only--not, however, if we go farther. The Aleutian for
_one_ = _attak-on_; the Aleutian for _six_ = _attu-on_. And what might be
the Aleutian for _seven_? Even _attakh-attun_, little more than _ittakh
tenu_ in a broader form.
The Jukahiri gives a similar phenomenon.
Such is the notice of the care with which certain comparisons should be
made before we venture to commit ourselves to negative statements.
There is an affinity amongst the American languages, and (there being
this) there are also the elements of a classification. The majority,
however, of the American languages must be classified according to
_types_ rather than _definitions_. Upon the nature of this difference,
as well as upon the cause I have written more fully elsewhere. It is
sufficient for present purposes to say that it applies to the languages
of North America in general, and (of these) to those of the parts
beyond the Rocky Mountains more especially. Eskimo characteristics
appear in the Athabaskan, Athabaskan in the Koluch forms of speech.
From these the Haidah leads to the Chimmesyan (which is, nevertheless,
a very outlying form of speech) and the Hailtsa, akin to the
Billechula, which, itself, leads to the Atna. By slightly raising the
value of the class we bring in the Kutani, the Nutkan and the Chinuk.
In the Chinuk neighbourhood we move _via_ the Jakon, Kalapuya,
Sahaptin, Shoshoni, and Lutuami to the languages of California and the
Pueblos; and thence southwards.
In American languages simple comparison does but little. We may test
this in two ways. We may place, side by side, two languages known to be
undoubtedly, but also known to be not very closely, allied. Such, for
instance, are the German and Greek, the Latin and Russian, the English
and Lithuanic, all of which are Indo-European, and all of which, when
placed in simple juxta-position, by no means show themselves in any
very palpable manner as such. This may be seen from the following
table, which is far from being the first which the present writer has
compiled; and that with the special view of ascertaining by induction
(and not _a priori_) the value of comparisons of the kind in question.