To isolate and study genes controlling floral development, we have carried out a large-scale transposon-mutagenesis experiment in Antirrhinum majus. Ten independent floral homeotic mutations were obtained that could be divided into three classes, depending on whether they affect (1) the identity of organs within the same whorl; (2) the identity and sometimes also the number of whorls; and (3) the fate of the axillary meristem that normally gives rise to the flower. The classes of floral phenotypes suggest a model for the genetic control of primordium fate in which class 2 genes are proposed to act in overlapping pairs of adjacent whorls so that their combinations at different positions along the radius of the flower can specify the fate and number of whorls. These could interact with class 1 genes, which vary in their action along the vertical axis of the flower to generate bilateral symmetry. Both of these classes may be ultimately regulated by class 3 genes required for flower initiation. The similarity between some of the homeotic phenotypes with those of other species suggests that the mechanisms controlling whorl identity and number have been highly conserved in plant evolution. Many of the mutations obtained show somatic and germinal instability characteristic of transposon insertions, allowing the cell-autonomy of floral homeotic genes to be tested for the first time. In addition, we show that the deficiens (def) gene (class 2) acts throughout organ development, but its action may be different at various developmental stages, accounting for the intermediate phenotypes conferred by certain def alleles. Expression of def early in development is not necessary for its later expression, indicating that other genes act throughout the development of specific organs to maintain def expression. Direct evidence that the mutations obtained were caused by transposons came from molecular analysis of leaf or flower pigmentation mutants, indicating that isolation of the homeotic genes should now be possible.