Much recent empirical work on hyperinflation has centered on the direct and indirect effects of uncertainty on the demand for money. We test the hypothesis, originally put forward by Klein, that uncertainty positively affects the demand for real balances. A variant of Cagan's demand for money function is utilized and operational measures of uncertainty are derived by fitting autoregressive integrated moving average models to the inflation series for each of three hyperinflation nations. We find that the most reasonable measures of variation do not significantly enter the money demand function for any of the three hyperinflations studied.