CRANV1P1 Test Design chapter Cyril Cleverdon Jack Mills Michael Keen Cranfield An investigation supported by a grant to Aslib by the National Science Foundation. Use, reproduction, or publication, in whole or in part, is permitted for any purpose of the United States Government. 18 in addition to the searches based on the Cranfield indexing, searches have also been made on the abstracts taken from the documents. This work is discussed in more detail in Chapter 7. It was, of course, known that decisions would have to be made concerning the physical methods which would be used for carrying out the searches. Fortunately, no firm decisions were taken on this point; the methods ultimately used are discussed in Chapter 6. Finally, there would be the necessity to present the results in a meaningful man- ner. The recall/precision ratio figures and curves of Cranfield I have undoubtedly taken a hammering over the past few years, and there are'many who have sought the elixir to change them into the pure gold of a single figure. Far from being able to do this, it was by 1961 clear to us that, if there was to be any comparison of ex- perimental results, it was necessary first to investigate the effect on performance of the generality ratio, namely the relationship between the number of relevant docu- ments and the size of the collection. The first tentative ideas on this had been put forward on page 101 of Ref. 2; in this project it was planned to attempt to measure the effect of this factor on recall and precision.