We implement an experimental design based on a duopoly game in order to analyse the impact of public subsidies on the willingness to cooperate in research and development (R&D) activities. We first implement six experimental markets by exogenously varying the level of knowledge spillovers (low or high) and the intensity of competition in the product market (low, intermediate, or high). We find that the probability of cooperation increases in the level of spillovers and decreases in that of market competition. We then replicate the six experimental markets by subsidising subjects who cooperate. Whenever they are sufficient to change the incentive structure of the game, subsidies substantially increase the probability of cooperation, causing, however, a reduction of R&D investments. Overall, our evidence suggests that, depending on the characteristics of the market, the provision of subsidies is not always desirable. These might be redundant because firms have sufficient private incentives to invest cooperatively in R&D, or even counterproductive, as they might induce firms to significantly reduce R&D investments compared to the noncooperative scenario.