We report on a simple experiment that addresses three factors in the frequency ofcooperation: (i) framing, (ii) the number of players and (iii) the perceived risk of cooperating. Wework with a (two-player) Prisoner s Dilemma and with a three-player, two-strategy Public GoodGame.These themes have been separately studied by a large number of experimental papers,using diverse methodologies. Our experiment targets them in a common, clear-cut framework. Wefind three strong effects (a) Framing; (b) The number of players (there is less cooperation in thethree-person games that in the two-person ones); (c) The neutralization of risk (about 50% ofparticipants cooperate when risk is neutralized). Both (a) and (c) go in the expected direction but,in all three cases, the strength of the effect is surprising.