This study examined the effects of criterion shift on false memory by manipulating payoff types and participants’ expectations about the probability of target words occurring in the recognition test. Participants in Experiment 1 received different sets of instructions that emphasized the negative consequence of either misses or false alarms. Results showed no changes in false recognition when the focus was on false alarm avoidance. In Experiment 2, the experimental groups studied the target words, whereas the control groups did not see any of the target words during the study phase. Both groups were also told that the probability of target words appearing in the recognition test was either one third or two thirds. Changes in the instructions did not affect performance of the experimental group, but had a large effect on performance of the control groups. Overall results of this study support the memory sensitivity view of false memory.