One of the few things we know about the present scientific communications network is that it is extraordinarily complicated; the useful network is certainly not confined to the conventional primary and secondary publications. Conferences and oral communications play an important part. Even sheer chance has a significant place. With this complexity, it is perhaps not surprising that we do not know very much about the efficiency with which the network is operating. An efficient system might be defined as one that ensures that all information likely to influence the programme of a research worker or research team is made available to them at the right time and in the most convenient form. Such a system should ensure that research is not delayed or conducted inefficiently because of the failure to obtain potentially available information, and that earlier research is not unnecessarily duplicated. Just to try and define efficiency indicates the difficulty of finding a way in which it can be measured quantitatively.