This paper reexamines evidence on M2 demand cointegration in the postwar United States. Equilibrium relations between M2 and various sets of its determinants are analyzed using quarterly observations from 1959:1 to 1988:4, 1990:4, and 1993:4 based on three different testing methods. For earlier subsamples, mixed results are obtained suggesting both cointegration and no cointegration. For the full sample, however, virtually no evidence supports cointegration. Accordingly, a M2 error-correction model is no longer appropriate to investigate short-run dynamics of business fluctuations in the 1990s. Copyright 1996 by Ohio State University Press.