Individual subsidy payments that are conditional on a collective contribution threshold could provide a viable resolution to the insufficient and dispersed adoption of agri-environmental contracts aiming at attaining environmental quality targets. Indeed, in a decontextualized laboratory experiment based on a threshold public good game (TPGG), Le Coent et al. (2014) offer promising results regarding a conditional subsidy compared to an unconditional subsidy (i.e. the standard subsidy in existing agrienvironmental schemes). In this article, we propose to improve the external validity of these results by transposing this laboratory experiment to the "field" with farmers. To do so, we carry out a contextualized lab-in-field experiment with farmers by explicitly mentioning agri-environmental contracts and water quality. Our results show that farmers cooperate even more successfully than students and sustain more efficient outcomes over time. Considering a between-subject, both in the lab with students, and in the "lab-in-field" experiment with farmers, our results indicate that average group contributions are not significantly different under the conditional subsidy mechanism and under the unconditional subsidy mechanism. Indeed the conditional subsidy introduces two behavioural responses (perceived risks and initial beliefs on others' contributions) having opposite effects on contributions, and compensating each other. The conditional incentive mechanism thus shows promising potential as a tool for agri-environmental policy since it avoids the pay-for-nothing trap of the unconditional subsidy mechanism without discouraging contributions.