This paper presents an experimental investigation of the three ambient-based mechanisms proposed by Segerson [J. Environ. Econom. Management 15, 87-98 (1988)] for controlling emissions from a group of nonpoint source polluters: a marginal tax/subsidy, a fixed penalty, and a mechanism that combines the two. To parallel likely conditions in a small watershed, in half of the experiment sessions we allow the group of polluters to engage in costless, non-binding discussion (cheap talk). In sessions without discussion, we find that the tax/subsidy instrument achieves the pollution target, while the fixed penalty and combined mechanisms do not. However, the tax/subsidy does not induce compliance at the individual level. Allowing discussion renders the tax/subsidy ineffective by encouraging gross overcompliance, but notably increases the usefulness of the fixed penalty and combined mechanisms. Differences in outcomes are likely attributable to the dissimilar marginal incentives under the three mechanisms.