Previously we have proposed a role for Bcl-3 in facilitating transactivation through kappa B sites by counteracting the inhibitory effects of bound, non-transactivating homodimers of the p50 subunit of NF-kappa B. Such homodimers are abundant for example in nuclei of unstimulated primary T cells. Here we extend the model and provide new evidence which fulfills a number of predictions. (i) Bcl-3 preferentially targets p50 homodimers over NF-kappa B heterodimers since the homodimers are completely dissociated from kappa B sites at concentrations of Bcl-3 which do not affect NF-kappa B. (ii) Select kappa B sites associate very strongly and stably with p50 homodimers, completely preventing binding by NF-kappa B. Such kappa B sites are likely candidates for regulation by p50 homodimers and Bcl-3. (iii) Bcl-3 and p50 can be co-localized in the nucleus, a requirement for active removal of homodimers from their binding sites in vivo. (iv) The ankyrin repeat domain of Bcl-3 is sufficient for the reversal of p50 homodimer-mediated inhibition, correlating with the ability of this domain alone to inhibit p50 binding to kappa B sites in vitro. Our data support the model that induction of nuclear Bcl-3 may be required during cellular stimulation to actively remove stably bound p50 homodimers from certain kappa B sites in order to allow transactivating NF-kappa B complexes to engage. This exact mechanism is demonstrated with in vitro experiments.