The Hammerstein identification problem is studied using a prediction error method in a separable least squares framework. Thus, the identification is recast as an optimization over the parameters used to describe the nonlinearity. Under certain conditions the identification problem is quasiconvex. First, the identification problem is shown to be quasiconvex under certain assumptions, including the use of an IID input. Next, the IID requirement is relaxed, and a sufficient condition for quasiconvexity is derived. The results are illustrated using a series of simulations.