Recent years have seen enticing empirical approaches to solving the epistemological problem of the theory-ladenness of observation. I group these approaches in two categories according to their method of choice: testing and refereeing. I argue that none deliver what friends of theory-neutrality want them to. Testing does not work because both evidence from cognitive neuroscience and perceptual pluralism independently invalidate the existence of a common observation core. Refereeing does not work because it treats theory-ladenness as a kind of superficial, removable (...) bias. Even if such treatment is plausible, there is likely no method to ascertain that effects of this bias are not present. More importantly, evidence from cognitive neuroscience suggests that a deeper, likely irremovable kind of theory-ladenness lies within the perceptual modules. (shrink)