A theory of best approximation with interpolatory contraints from a fi
nite-dimensional subspace M of a normed linear space X is developed. I
n particular, to each x is an element of X, best approximations are so
ught from a subset M(x) of M which depends on the element x being appr
oximated. It is shown that this ''parametric approximation'' problem c
an be essentially reduced to the ''usual'' one involving a certain fix
ed subspace M(0) of M. More detailed results can ve obtained when (1)
X is a Hilbert space, or (2) M is an ''interpolating subspace'' of X (
in the sense of [1]). (C) 1996 Academic Press, Inc.