We explore a meta-indexing scheme that we have called {\bf n-term} indexing, to solve the {\bf collection selection} problem. For a given query, a collection selection algorithm should rank a set of known document collections in order of user-relevance. Several selection algorithms employ lexicons that record every unique term from the set of indexed documents; such schemes limit the amount of other information that may be stored with the indexed terms, and therefore potentially limiting the retrieval effectiveness of the algorithms. An n-term index is a meta-indexing scheme that captures $n$ representative for each document in each collection. Such a scheme may be adapted to suit the characteristics of the federated database, through judicious choice of $n$. This paper presents issues, results and future plans, relating to our interest in n-term indexing, to solve the collection selection problem.