We present in detail how a statistical analysis of variance (ANOVA) is used to sort out the effect of an unexpected batch-to-batch variation between cell cultures. Two separate cultures of rainbow trout branchial cells were grown on permeable filtersupports ("inserts"). They were supposed to be simple duplicates for testing the effect of two induced factors-apical or basolateral addition of radioactive precursors and different apical media-on the incorporation of 14C-acetate and 32Pphosphate intotissue lipids. Unfortunately, they did not altogether give the same result. By accepting this fact and introducing the observed difference between batches as one of the factors in an expanded three-dimensional ANOVA, we were able to overcome an otherwisecrucial lack of sufficiently reproducible duplicate values. We could thereby show that the effect of changing the apical medium was much more marked when the radioactive lipid precursors were added on the apical, rather than on the basolateral, side. Theinsert cell cultures were obviously polarized. We argue that it is not reasonable to reject troublesome experimental results, when we do not know a priori that something went wrong. The ANOVA is a very useful statistical tool that can be utilized by anyexperimentalist who cares to make the effort.