Thanks for the reply.

My example was poorly chosen since it does look as though I'm trying to
calculate the sum of squares.

I'm really trying to check that a set of eigenvectors are orthonormal.
I'm aiming to calculate

=SUMPRODUCT(X(i),X(j)) (via =SUM(X(i)*X(j)) )

where i,j = 1, ..., 31, and X(i) is a 31x1 column vector of real
values. The calculations are to be placed in a 31x31 matrix which
should be the identity matrix.

I have the eigenvectors in adjacent columns, so to pick them out as I
create the matrix, I'm using X(1) plus a column offset relative to the
first eigenvector.

Hope this helps to explain the situation. I'm certainly not saying it
is the most efficient approach.

Thanks again,

Clive