In order to quantify higher-order correlations of the galaxy cluster distribution we use a complete family of additive measures which give scale-dependent morphological information. Minkowski functionals can be expressed analytically in terms of integrals of n-point correlation functions. They can be compared with measured Minkowski functionals of volume limited samples extracted from the REFLEX survey. We find significant non-Gaussian features in the large-scale spatial distribution of galaxy clusters. A Gauss-Poisson process can be excluded as a viable model for the distribution of galaxy clusters at the significance level of 95%.