Shariff and Norenzayan (2007) discovered that people allocate more money to anonymous strangers in a dictator game following a scrambled sentence task that involved words with religious meanings. We conducted a direct replication of key elements of Shariff and Norenzayan’s (2007) Experiment 2, with some additional changes. Specifically, we (a) collected data from a much larger sample of participants (N = 650); (b) added a second religious priming condition that attempted to prime thoughts of religion less conspicuously; (c) modified the wording of some of their task explanations to avoid deceiving our participants; (d) added a more explicit awareness probe; (e) reduced prime-probe time; and (f) performed statistical analyses that are more appropriate for non-normal data. We did not find a statistically significant effect for religious priming. Additional tests for possible between-subjects moderators of the religious priming effect also yielded nonsignificant results. A small-scale meta-analysis, which included all known studies investigating the effect of religious priming on dictator game offers, suggested that the mean effect size is not different from zero, although the wide confidence intervals indicate that conclusions regarding this effect should be drawn with caution. Finally, we found some evidence of small-study effects: Studies with larger samples tended to produce smaller effects (a pattern consistent with publication bias). Overall, these results suggest that the effects of religious priming on dictator game allocations might be either not reliable or else quite sensitive to differences in methods or in the populations in which the effect has been examined.