VAR studies of the effects of monetary policy on output suggest that a contractionary impulse results in a drawn-out, hump-shaped response of output. Standard structural economic models are generally not able to reproduce such a response. In this paper I look at nonfundamental representations that are observationally equivalent to a VAR. I find that the quantitative effect of a monetary policy shock on output might be much smaller and much more short-lived than the VAR studies suggest. I conclude that the apparent discrepancy between the VAR findings and standard structural models may be spurious and that the general tendency to append non-structural, ad hoc features to structural models should be questioned.