Why count(*) doest use index?

I use pgsql 9.0.3 and I know that postgresql tries to use the fields
in indexes instead of the original table if it possible

But when I run

SELECT COUNT(id) FROM tab

or

SELECT COUNT(*) FROM tab

where
there "id" is PRIMARY KEY and there are other indexes there I get
execution plan that doesnt use any indexes, but sequentab scanning the
original table.

"Aggregate (cost=38685.98..38685.99 rows=1 width=0)"

" -> Seq Scan on tab (cost=0.00..36372.38 rows=925438 width=0)"

Why is
it so?

---

Paul