We present a technique for improving retrieval performance which we call clipping. Clipping is based on the manipulation of the original query terms, without attempting query expansion. The idea is to use a small set of query terms, which we call a clipping set, to clip the original answer set reducing the number of documents in it. We discuss the alternatives that such clipping technique opens up and characterize what we call the clipping problem. This problem consists basically of finding a good clipped answer set (i.e., one which yields improved retrieval performance) among many possible ones. We show that, for queries with a proper formulation, gains in precision as high as 50% can be obtained.