Convective parameterizations used in general circulation models (GCMs) generally only simulate the mean or first-order moment of convective ensembles and do not explicitly include higher-order moments. The influence of including unresolved higher-order moments is investigated using a simple stochastic convective parameterization that includes a random contribution to the convective available potential energy (CAPE) in the deep convective scheme. Impacts are tested in a tropical atmospheric model of intermediate complexity. Adding convective noise noticeably affects tropical intraseasonal variability, suggesting inclusion of such noise in GCMs might be beneficial. Model response to the noise is sensitive not only to the noise amplitude, but also to such particulars of the stochastic parameterization as autocorrelation time. ¿ 2000 American Geophysical Union