AbstractProblems, specific for QSAR studies of steroids, are reviewed. The attention is drawn to the extensive use of X‐ray structure determinations to assist in weeding‐out chance correlations which arise from the need to use too small series of derivatives and the tendency in older studies to underestimate the flexibility of steroids.A series of papers on the QSAR of the cardiac glycosides illustrate some problems associated with the use of molecular modelling.