We examined the extent to which transfer of training is impacted by target probability and physical similarity of training and transfer stimuli. Participants performed a luggage screening task where they trained on three base rates of weapon presence as follows: 100%, 50% and 20%; at transfer, participants detected the same weapons (Study 1) or novel weapons (Studies 2 and 3) at a base rate of 20%. In Study 3, we modified training such that all participants observed the same absolute number of weapons (n = 100) across base rate conditions. In Study 1, different base rates did not significantly impact transfer. In Study 2, 100% base rates helped hit rates at transfer but increased false alarms via liberal shifts in decision criteria and slowed detection times. Study 3 revealed no effect of base rates on transfer hit rates; however, training on 100% base rates still led to liberal responding and inflated false alarms. Copyright © 2013 John Wiley & Sons, Ltd.