We conduct field and laboratory experiments on the same panel of experts, measuring the internal consistency of their predictions 1) in the field, in their domain of expertise, and 2) on a conceptually identical laboratory exercise. Experts make internally consistent predictions in the field, both in absolute terms and relative to a panel of novices, but they exhibit markedly less consistency on the isomorphic lab exercise. Possible explanations for this fading expertise include low motivation in the lab and a failure to transfer skills learned implicitly in the field to the more abstract lab setting.