Guidant Corporation has been generous with information, helping in early comparisons of Monte Carlo calculations and more recently in discussions about possible modeling artifacts. Paul Bergstrom of NIST was of invaluable help in setting up and running the comparison calculations with the MCNP4C, EGS4, EGSnrc, and PENELOPE codes for this manuscript, on very short notice and with a very short deadline. And my thanks to John DeMarco and Dennis Schaart who provided pre-print material that helped in my better understanding of the MCNP code. 20 It can no doubt be argued that improved point-kernel models are possible. It would seem, however, that pointkernel approaches could not reproduce dose perturbations close to the interface separating two materials of significantly different atomic numbers, a situation not uncommon in these applications.