The floral-dip method for Agrobacterium-mediated transformation of Arabidopsis allows efficient plant transformation without need for tissue culture. To facilitate use with other plant species, we investigated the mechanisms that underlie this method. In manual outcrossing experiments, application of Agrobacterium tumefaciens to pollen donor plants did not produce any transformed progeny, whereas application of Agrobacterium to pollen recipient plants yielded transformants at a rate of 0.48%. Agrobacterium strains with T-DNA carrying gusA (encoding beta-glucuronidase [GUS]) under the control of 35S, LAT52, or ACT11 promoters revealed delivery of GUS activity to developing ovules, whereas no GUS staining of pollen or pollen tubes was observed. Transformants derived from the same seed pod contained independent T-DNA integration events. In Arabidopsis flowers, the gynoecium develops as an open, vase-like structure that fuses to form closed locules roughly 3 d prior to anthesis. In correlation with this fact, we found that the timing of Agrobacterium infection was critical. Transformants were obtained and GUS staining of ovules and embryo sacs was observed only if the Agrobacterium were applied 5 d or more prior to anthesis. A 6-fold higher rate of transformation was obtained with a CRABS-CLAW mutant that maintains an open gynoecium. Our results suggest that ovules are the site of productive transformation in the floral-dip method, and further suggest that Agrobacterium must be delivered to the interior of the developing gynoecium prior to locule closure if efficient transformation is to be achieved.