We apply a method for quantifying the topology of projected galaxy clustering to the Abell and ACO catalogues of rich clusters. We use numerical simulations to quantify the statistical bias involved in using high peaks to define the large-scale structure, and we use the results obtained to correct our observational determinations for this known selection effect and also for possible errors introduced by boundary effects. We find that the Abell cluster sample is consistent with clusters being identified with high peaks of a Gaussian random field, but that the ACO shows a slight meatball shift away from the Gaussian behaviour over and above that expected purely from the high-peak selection. This effect is of low statistical significance but seems consistent with other quite separate statistical analyses of the ACO sample. The most conservative explanation of this effect is that it is caused by some artefact of the procedure used to select the clusters in the two samples. On the other hand, our results could be indicating that there is either a large-scale variation in the cluster distribution due to superclustering on scales > 100 h^-1^ Mpc, or that there is some intrinsic non-Gaussian signature in the cluster distribution which is masked in the Abell samples by projection contamination.