We have reanalyzed a data set of 99 low redshift ($ z < 0.1 $) Abell clusters and determined their shapes. For this, three different measures are used. We use Monte-Carlo simulations to investigate the errors in the methods. The corrected distribution of cluster ellipticities shows a peak at $ ��\sim 0.4 $ and extends to $ ��\sim 0.8 $. The results are self-consistent, i.e., with the corrected distribution over projected cluster shapes we can reconstruct the observed distribution and the observed relation between the number of galaxies in a cluster and its ellipticity. It is shown that the richer clusters are intrinsically more nearly spherical than the poorer ones. The corrected distribution of cluster shapes is more consistent with a prolate population than with an oblate population. We compare the corrected true distribution of (projected) ellipticities with predictions from N-body simulations. For this, we use a catalogue of 75 N-body simulated clusters (van Kampen 1994) which assume a CDM spectrum with $ ��= 1.0 $. The simulations include a recipe for galaxy formation and merging. 'Observing' these simulated clusters produces an ellipticity distribution that extends to much higher $ ��$ and that has too few nearly spherical clusters. Preliminary results of simulations of the formation of clusters in an $ ��= 0.2 $ universe suggest that, on average, clusters are more nearly spherical in this case.

17 pages, postscript, submitted to MNRAS