We use direct-summation N-body integrations to follow the evolution of binary black holes at the centers of galaxy models with large,
constant-density cores. Particle numbers as large as 0.4x106 are considered.
The results are compared with the predictions of loss-cone theory, under the assumption that the supply of stars to the binary is
limited by the rate at which they can be scattered into the binary's influence sphere by gravitational encounters.
The agreement between theory and simulation is quite good; in particular, we are able to quantitatively explain the observed
dependence of binary hardening rate on N. We do not verify the recent claim of Chatterjee, Hernquist & Loeb (2003) that the
hardening rate of the binary stabilizes when N exceeds a particular value, or that Brownian wandering of the binary has a
significant effect on its evolution. When scaled to real galaxies, our results suggest that massive black hole binaries in
gas-poor nuclei would be unlikely to reach gravitational-wave coalescence in a Hubble time.