Three recent papers measure the marginal excess burden of labor taxes in the United States. They obtain very different results even where they all use a zero uncompensated labor supply elasticity and assume that the additional revenue is spent on a public good that is separable in utility. The impression is that other parameters must explain the differences in results. Yet each paper uses a different concept of excess burden. Here, I calculate all three measures in one model and show how conceptual differences explain the results. Only one of these measures isolates the distortionary effects of taxes in a way that depends on the compensated labor supply elasticity. The other two measures incorporate income effects and thus depend on the actual change in labor. This result was obscured because those papers report positive marginal excess burden even with a zero uncomspensated labor supply elasticity. This paper shows conditions under which their measure is zero, and it interprets the measures in light of recent literature.