Many search strategies have been proposed for use in information retrieval systems, but none is clearly dominant. A strategy that works well for one query may not work as well for another. In view of this, it would be desirable to be able to select an appropriate search strategy automatically. This paper describes an investigation into the use of statistical measures to select either a clustered or a nonclustered search strategy in a best-match retrieval system. Four test collections were used, and nine measures based on query terms, query-document similarities, and document-document similarities were considered. None of the individual measures was found to be effective in selecting between the two strategies; combinations of measures were not considered. This paper presents important negative results and gives additional support to the tentative conclusion that statistical measures cannot effectively be used to select among search strategies.