A measure of association for spatial variables is proposed. The measure is based on the ranks of the observations and on the location coordinates of the measurement points. For two independent spatial variables it is shown that under optimal coordinate designs the variance of the measure is of the order l/(kn), where n is the number of data points and k is the dimension of the observation space. Conditions for asymptotic normality are stated, and asymptotic formulae for bias and variance are given also in the case of sampling from a population with a finite number of measurement points.