Drawing on an evaluation of past and current risks to light water power reactors arising from seismic design and construction errors, as well as the results of recent probabilistic risk assessments, a historical comparison is made between expert estimates of seismic risk and the implications of experience with past seismic deficiencies. A few methods of annual Bayesian updating of expert opinion on seismic risk as a function of new information are compared. The implications of the results include the suggestions that care be exercised in using a self-estimate of uncertainty in performing weighting and that the use of common information may lead consciously or unconsciously to a disadvantageous excessive weight placed on prevalent rather than innovative opinion.