Most studies of genetic interactions have measured a single phenotypic readout (usually growth rate in the case of Saccharomyces cerevisiae) under one set of experimental conditions, but several studies have recently generated networks using different phenotypes and conditions. This paper presents a systematic analysis of such data from Saccharomyces cerevisiae to assess the extent to which interaction networks vary. The authors show that using a greater range of conditions will increase insights from interaction studies, and they describe a method for helping to combine the networks that are generated.