AbstractYield data from 5 rotation cycles in the long-term soil fertility experiments in Skåne were submitted to analysis of variance. Each crop was analysed separately. Only the mean yield over blocks were available for analysis and, instead of blocks, fields were considered as replicates. Years were considered to be a fixed factor. Most treatment effects and treatment interactions were found to be significant. Some linear trends were also significant. Alternative models for analysing the data are discussed.