ABSTRACT
Science journal managers should exercise care in preparing data for testing correlations between use and citation data. Correlations should be sought only among journals of fairly similar subject specialty, scope, purpose, and language rather than among journals in a broad field, e.g., science overall. Either gross citation ranking or impact factor will usually correlate well with use, except in cases where a journal is either new or characteristically publishes a few papers. In these cases impact factor must be used in comparisons. In order for the comparisons to have statistical validity there should be relatively heavy overall use, an average of 25 potential borrowings per title in the subject specialty being analyzed. The authors present tables showing good correlations when these conditions are met and other tables showing poor correlations are analyzed in terms of unmet conditions.
