The sizes and shapes of voids in a galaxy survey depend not only on the phy
sics of structure formation, but also on the sampling density of the survey
and on the algorithm used to define voids. Using an N-body simulation with
a tau CDM power spectrum, we study the properties of voids in samples with
different number densities of galaxies, in both redshift space and real sp
ace. When voids are defined as totally empty regions of space, their charac
teristic volume is strongly dependent on sampling density; when they are de
fined as regions whose density is 0.2 times the mean galaxy density, the de
pendence is less strong. We compare two void-finding algorithms, one in whi
ch voids are nonoverlapping spheres, and one, based on the algorithm of Aik
io & Mahonen, that does not predefine the shape of a void. Regardless of th
e algorithm chosen, the characteristic void size is larger in redshift spac
e than in real space, and is larger for low sampling densities than for hig
h sampling densities. We define an elongation statistic Q that measures the
tendency of voids to be stretched or squashed along the line of sight. Usi
ng this statistic, we find that at sufficiently high sampling densities (co
mparable to the number density of L > L-* galaxies), large voids tend to be
slightly elongated along the line of sight in redshift space.