At first the individuals composing the pair are simply "this one," and "that one," or "this and that"; and his number system now halts for a time at the stage when he can, rudely enough it may be, count 1, 2, many.

There are certain cases where the forms of 1 and 2 are so similar than one may readily imagine that these numbers really were "this" and "that" in the savage's original conception of them; and the same likeness also occurs in the words for 3 and 4, which may readily enough have been a second "this" and a second "that." In the Lushu tongue the words for 1 and 2 are _tizi_ and _tazi_ respectively.

In Koriak we find _ngroka_, 3, and _ngraka_, 4; in Kolyma, _niyokh_, 3, and _niyakh_, 4; and in Kamtschatkan, _tsuk_, 3, and _tsaak_, 4.[108] Sometimes, as in the case of the Australian races, the entire extent of the count is carried through by means of pairs.

But the natural theory one would form is, that 2 is the halting place for a very long time; that up to this point the fingers may or may not have been used--probably not; and that when the next start is made, and 3, 4, 5, and so on are counted, the fingers first come into requisition.

If the grammatical structure of the earlier languages of the world's history is examined, the student is struck with the prevalence of the dual number in them--something which tends to disappear as language undergoes extended development.