Prosocial behaviour in chimpanzees is a topic of much debate. Earlier studies with game theoretical approaches suggest chimpanzees to be rational maximizers without other-regarding preferences. Here, we present data from two studies with a special emphasis on ensuring that subjects understood the payoff structures and the consequences of their decisions before the experiment. On a table between two cages, there were three initially inaccessible payoff options that could allocate food bowls to both sides. From side 1, either the first of these payoffs could be chosen or it could be discarded and the choice between the other two payoffs moved over to side 2. Six of twelve subjects passed several steps of criteria that ensured complete understanding of the apparatus and set-up. These six subjects were then paired with a conspecific partner on side 1 that was trained to always pass over the choice to them. In study 1, the partner discarded a 2/0 payoff (two food pieces for oneself, none for the other one) and provided the subjects with the choice between a 2/0 and a 2/2 payoff. If subjects were self-regarding rational maximizers, they should then have chosen randomly. However, they chose the prosocial option significantly above chance level (~75% of trials). In contrast, in a control condition in which the experimenter rather than the conspecific allowed them to choose they did so at chance level. Thus, chimpanzees provided benefits for a conspecific after they had received assistance from her but did not act prosocially in general. In the ongoing study 2, we introduced costs to providing benefits to the partner. The partner discarded the payoff of 3/0 and allowed the subjects to choose between 4/0 and 3/3. Preliminary results suggest a marked drop in prosocial choices but still a difference between the test and control condition.