To get more intuition for working with matrices, it might also help to consider a $n$ x $m$ matrix A as a representation of a linear transformation from an m-dimensional vector space $V$ to an n-dimensional vector space $W$, see: http://en.wikipedia.org/wiki/Linear_transformation
A is not just a table of numbers with related operations; it produces elements of $W$ when fed with elements of $V$.
If you feed A with one $V$ element, it produces one $W$ element; if you feed it with two, it produces two. Also note that the linearity of the transformation implies (and is defined by) that for $v_1, v_2 \in V$ and $\lambda, \mu$ scalars, A fed with $\lambda v_1 + \mu v_2$ will produce $\lambda Av_1 + \mu Av_2$, thus: $A(\lambda v_1 + \mu v_2) = \lambda A(v_1) + \mu A(v_2)$