We have reanalysed a data set of 99 low-redshift (z < 0.1) Abell clust
ers studied previously by Rhee, van Haarlem & Katgert, and determined
their shapes. For this, three different measures are used: two of whic
h were originally used by Rhee et al., and one of which was used by Pl
ionis, Barrow & Frenk in their investigation of clusters in the Lick c
atalogue. We use Monte Carlo simulations of clusters to investigate th
e errors in the methods. For low ellipticity, all methods overestimate
the cluster elongation, whereas the opposite is true for a highly fla
ttened system. Also background galaxies and shot noise have a rather l
arge influence on the measured quantities. The corrected distribution
of cluster ellipticities shows a peak at epsilon similar to 0.4 and ex
tends to epsilon similar to 0.8, consistent with results of some previ
ous studies. However, the present study uses more than twice as many c
lusters as the earlier studies, and is self-consistent. That is, with
the corrected distribution over projected cluster shapes we can recons
truct the observed distribution over projected cluster shapes and the
observed relation between the number of galaxies in a cluster and its
ellipticity. To achieve this, we have to assume that there is an anti-
correlation between the true (projected) ellipticity of a cluster and
its number of galaxies. It is not necessary to assume that the ellipti
city of a cluster increases when one only includes the brighter galaxi
es (as suggested by Binney). Using a redshift-independent richness cri
terion of Vink & Katgert, it is shown that the richer clusters are int
rinsically more nearly spherical than the poorer ones. Furthermore, th
e corrected distribution of cluster shapes is found to be more consist
ent with a population that consists of purely prolate clusters than wi
th a purely oblate population. We compare the corrected true distribut
ion of (projected) ellipticities with predictions from N-body simulati
ons. For this, we use a catalogue of 75 N-body simulated clusters whic
h assume a CDM spectrum with Omega = 1.0. The simulations include a re
cipe for galaxy formation and merging. The model clusters are expected
to be a representative sample of all real clusters. They show a good
resemblance with the data in both radial profile and number of galaxie
s. 'Observation' of these simulated clusters in exactly the same way a
s the real clusters produces an ellipticity distribution that extends
to much higher epsilon and that has too few nearly spherical clusters.
Preliminary results of simulations of the formation of clusters in an
Omega = 0.2 universe suggest that, on average, clusters are more near
ly spherical in this case, as is expected on theoretical grounds. This
shows that the elongations of clusters can provide a useful constrain
t on the value of Omega.