We analyze dictator allocation decisions in an experiment where the recipients have to earn the pot to be divided with a real-effort task. As the recipients move before the dictators, their effort decisions resemble the first move in a trust game. Depending on the recipients' performance, the size of the pot is either high or low. We compare this real-effort treatment to a baseline treatment where the pot is a windfall gain and where a lottery determines the pot size. In the baseline treatment, reciprocity cannot play a role. We find that female dictators show reciprocity and decrease their taking-rates significantly in the real-effort treatment. This treatment effect is larger when female dictators make a decision on recipients who successfully generated a large pot compared to the case where the recipients performed poorly. By contrast, there is no treatment effect with male dictators, who generally exhibit more sefish behavior.