The WSU-UX algorithm, a natural choice for incentive-compatible online learning with bandit feedback, cannot achieve regret better than Ω(T^2/3) in the worst case.