We design a controlled experiment to examine investment behavior when a proportional income tax (with and without full loss offset) is introduced. Our design enables us to disentangle tax effects predicted by standard neoclassical theory and behavioral tax biases that cannot be predicted by this theory. We observe that investment behavior can be heavily distorted by behavioral tax biases and is consequently inconsistent with theoretical predictions. However, if we isolate these effects, we find support for the theory. We argue that the unpredicted behavior isa result of a tax-induced cognitive load variation.