We show that if policy-makers compute the optimal unconstrained interest-rate rule within a Taylor-type class, they may be led to rules that generate indeterminacy and/or instability under learning. This problem is compounded ...