```Christophe Grandsire scripsit:

> [I]f a French cook asked me to put "a little" butter in a pan, I would put just
> enough to make the pan slightly greasy, and I find that normal. On the other
> hand, when Jan asks me to put "a little" butter in the pan, he expects me to
> put about 50g of butter,

For Anglo-Saxon cooks, that is about 3 1/2 tbsp, or just under half
a stick.  Odd that you measure butter by mass, but we measure it by
volume: 1 tbsp = 14.8 cm^3.  Luckily, the GNU units program knows the
density of butter: 958.61142 kg / m^3.

(Other sources claim 860 and 800 kg / m^3, alas.  To heck with it.)

>  [T]he actual value of quantifiers depend on the
> language (sometimes on the person who speaks it too, but the language
> dependence is strong).

Other examples:  Polish "orange" covers fewer shades than English "orange".
Child "soon" is much sooner than adult "soon", at least in English.

> So to bring back the subject to conlang matters, I've been wondering how you
> people thought about the quantifiers in your conlangs, and if you actually
> thought of those problems of presupposed value of a "little" in your conlangs.

In Lojban, there are five vague quantifiers, usually translated "almost all",
"most", "many", "several", and "a few".  But all that is really defined is
that if more than one is used in the same context, they form a scale in
the order given.  There is no implication, for example, that "most" means
"a majority" (more than half).

It is always possible to add a case tag meaning "in-system" and specify
a context to get these things exactly right.

--