One of the key phenomena in the adaptive immune response to infection and immunization is affinity maturation, during which antibody genes are mutated and selected, typically resulting in a substantial increase in binding affinity to the eliciting antigen. Advances in technology on several fronts have made it possible to clone large numbers of heavy-chain light-chain pairs from individual B cells and thereby identify whole sets of clonally related antibodies. These collections could provide the information necessary to reconstruct their own history - the sequence of changes introduced into the lineage during the development of the clone - and to study affinity maturation in detail. But the success of such a program depends entirely on accurately inferring the founding ancestor and the other unobserved intermediates. Given a set of clonally related immunoglobulin V-region genes, the method described here allows one to compute the posterior distribution over their possible ancestors, thereby giving a thorough accounting of the uncertainty inherent in the reconstruction. I demonstrate the application of this method on heavy-chain and light-chain clones, assess the reliability of the inference, and discuss the sources of uncertainty.