This paper examines how individual adoption decisions of agricultural technologies depend on the choices of others in social networks. Using a unique dataset on sunflower adoption in Northern Mozambique, the study finds that the relationship between the probability of adoption and the number of known adopters is shaped as an inverse-U. Farmers are more likely to adopt when they know few other adopters, but this probability decreases as they know more adopters. This suggests a trade-off between static (positive) and dynamic (negative) network effects. The study also finds that information sharing is an important channel through which group behavior affects individual adoption choices, with network effects being stronger for farmers who discuss agriculture with others. The data allows for precise identification of social networks and control for village heterogeneity and endogenous group formation. The paper contributes to the literature by introducing a new dataset that addresses identification issues and by presenting evidence on the precise shape of the relationship between individual adoption decisions and group behavior. The results suggest that existing estimates are likely to be downward biased because non-linearities are not taken into account and the social network is not precisely measured. The study also addresses econometric concerns such as correlated unobservables and simultaneity, and shows that the results are robust to these concerns. The findings suggest that information sharing is an important determinant of adoption decisions, and that social effects do matter. However, the study cannot separate endogenous social effects from exogenous social effects.This paper examines how individual adoption decisions of agricultural technologies depend on the choices of others in social networks. Using a unique dataset on sunflower adoption in Northern Mozambique, the study finds that the relationship between the probability of adoption and the number of known adopters is shaped as an inverse-U. Farmers are more likely to adopt when they know few other adopters, but this probability decreases as they know more adopters. This suggests a trade-off between static (positive) and dynamic (negative) network effects. The study also finds that information sharing is an important channel through which group behavior affects individual adoption choices, with network effects being stronger for farmers who discuss agriculture with others. The data allows for precise identification of social networks and control for village heterogeneity and endogenous group formation. The paper contributes to the literature by introducing a new dataset that addresses identification issues and by presenting evidence on the precise shape of the relationship between individual adoption decisions and group behavior. The results suggest that existing estimates are likely to be downward biased because non-linearities are not taken into account and the social network is not precisely measured. The study also addresses econometric concerns such as correlated unobservables and simultaneity, and shows that the results are robust to these concerns. The findings suggest that information sharing is an important determinant of adoption decisions, and that social effects do matter. However, the study cannot separate endogenous social effects from exogenous social effects.