Theory typically does not give us reason to believe that economic models ought to be formulated at the same level of time aggregation at which data happen to be available. Nevertheless, this is frequently done when formulating econometric models, with potentially important specification-error implications. This suggests examining the alternatives, one of which is to model in continuous time. The primary difficulty in inferring the parameters of a continuous time model given sampled observations is the “aliasing identification problem.” This paper shows how the restrictions implied by rational expectations sometimes do, and sometimes do not, resolve the problem. This is accomplished very simply in the context of a hypothesis about the term structure of interest rates. The paper confirms and extends results obtained for another example by Hansen and Sargent.