We present an analysis of the clustering evolution of dark matter in f
our cold dark matter (CDM) cosmologies. We use a suite of high-resolut
ion, 17 million particle, N-body simulations that sample volumes large
enough to give clustering statistics with unprecedented accuracy. We
investigate a flat model with Omega(0) = 0.3, an open model also with
Omega(0) = 0.3, and two models with Omega = 1, one with the standard C
DM power spectrum and the other with the same power spectrum as the Om
ega(0) = 0.3 models. In all cases, the amplitude of primordial fluctua
tions is set so that the models reproduce the observed abundance of ri
ch galaxy clusters by the present day. We compute mass two-point corre
lation functions and power spectra over 3 orders of magnitude in spati
al scale and find that in all of our simulations they differ significa
ntly from those of the observed galaxy distribution, in both shape and
amplitude. Thus, for any of these models to provide an acceptable rep
resentation of reality, the distribution of galaxies must be biased re
lative to the mass in a nontrivial, scale-dependent fashion. In the Om
ega = 1 models, the required bias is always greater than unity, but in
the Omega(0) = 0.3 models, an ''antibias'' is required on scales smal
ler than similar to 5 h(-1) Mpc. The mass correlation functions in the
simulations are well fit by recently published analytic models. The v
elocity fields are remarkably similar in all the models, whether they
are characterized as bulk flows, single-particle, or pairwise velocity
dispersions. This similarity is a direct consequence of our adopted n
ormalization and runs contrary to the common belief that the amplitude
of the observed galaxy velocity fields can be used to constrain the v
alue of Omega(0). The small-scale pairwise velocity dispersion of the
dark matter is somewhat larger than recent determinations from galaxy
redshift surveys, but the bulk Bows predicted by our models are broadl
y in agreement with most available data.