THEORY: Relation between counting, trial, and plural

In Ainu, days are counted in a way which is strange to me. Here are
the basic morphemes for the numbers 1-5:
"one"	sine
"two"	tu
"three"	re
"four"	ine
"five"	asikne

Here is a summary of counting days from 1 to 5:

"one day"	_sine to_
"two days"	_tutko_
"three days"	_rerko_
"four days"	_ine rerko_
"five days"	_asikne rerko_

There are a few interesting things here. First, the word for "day(s)"
seems to have an allomorph _to_ for just one day, and _ko_ for other
numbers of days. Second, _tutko_ and _rerko_ look as though they
might be formed by reduplicating the forms for "two" and "three" and
then adding _ko_. Now, what I find really interesting is how you say
"four (or more) days" -- you use a numeral plus the word which (by
itself) means "three days"!

Are there any other natlang examples of things like this? I could
have sworn that I used to know of one, but I've forgotten it.

Are there any (other) natlangs which have "conflate" trial and plural
-- which have a grammatical number which isn't quite trial and isn't
quite plural -- i.e. where words marked with that number are assumed
to refer to three things, unless specifically quantified? Further, do
any natlangs conflate *dual* and plural in that way?

Replies