We analysed the distribution of void sizes in the two-dimensional slices of
the Las Campanas Redshift Survey (LCRS). Fourteen volume-limited subsample
s were extracted from the six slices to cover a large part of the survey an
d to test the robustness of the results against cosmic variance. Thirteen s
amples were randomly culled to produce homogeneously selected samples. We t
hen studied the relationship between the cumulative area covered by voids a
nd the void size as a property of the void hierarchy. We found that the dis
tribution of void sizes scaled with the mean galaxy separation, lambda. In
particular, we found that the size of voids covering half of the area is gi
ven by D-med approximate to lambda + (12 +/- 3) h(-1) Mpc. Next, by employi
ng an environmental density threshold criterion to identify mock galaxies,
we were able to extend this analysis to mock samples from dynamical N-body
simulations of cold dark matter (CDM) models. To reproduce the observed voi
d statistics, overdensity thresholds of delta(th) approximate to 0, ..., 1
are necessary. We compared standard (SCDM), open (OCDM), vacuum energy domi
nated (Lambda CDM) and broken scale invariant CDM models (BCDM): we found t
hat both the void size distribution and the two-point correlation function
provided important and complementary information on the large-scale matter
distribution. The dependence of the void statistics on the threshold criter
ion for the mock galaxy identification showed that the galaxy biasing was m
ore crucial for the void size distribution than were differences between th
e cosmological models.