The choice of environmental taxes is one of the central themes in policy design for agricultural pollution control, which dominates both empirical and theoretical research. This paper examines two assumptions very often employed in applied research, namely the absence of transaction costs and the normality of agricultural pollutants. Our results indicate that the well-known superiority of emission taxes over input taxes may not always be valid, when transaction costs are taken into account. Further more, the assumption of normal distributed agricultural pollutants overestimate the relative abatement costs.