We have begun work on how to use a large corpus of literature on a topic to most efficiently extend the set of documents in such a database by related material. This is fundamentally different than the problem of how to find a document closely related to a single document. One is given much more data. Thus far we have used a logodds approach with some success. Another approach which we hope to try in the future is the pooling of neighborhood data based on single documents in the database. Such pooling may give a good idea of what is likely to be useful material. This whole problem of extension can also be studied from the point of view of the initial construction of a literature. We hope to look at this problem also and propose to measure the success of a method by the number documents that must be examined in order to find a given number of relevant documents.