Geochemical and petrologic data from contact aureoles consistently document fluid focusing through small-scale permeable structures. We use stochastic representations of permeability in a series of transient numerical simulations to assess how such small-scale rock heterogeneities influence kilometer-scale fluid convection around a shallow crustal pluton. The sensitivity study considers different permeability scenarios by varying statistical characteristics of the permeability distribution (mean, variance, and spatial correlation). Large-scale convective flow patterns in heterogeneous contact aureoles are shown to deviate significantly from equivalent homogeneous aureoles in ways that cannot be predicted without detailed mapping of the permeability field. Fluid focused through high-permeability zones results in cumulative fluid fluxes as high as 1¿108 kg m-2 after 2¿105 years, while nearby low-permeability regions experience time-integrated fluxes ≪105 kg m-2. Spatially variable flow patterns preclude extrapolation of flow directions or fluxes determined from one portion of the aureole to another. Several simulations show high-permeability zones developing isolated long-lived convection cells far from the intrusion. Reduced permeability contrasts (low variance) or anisotropic permeabilities inhibit development of such hydrologically isolated regions. When average aureole permeability is 10-16 m2, maximum metamorphic temperatures deviate by ≤30¿ from conductive cooling models despite locally high fluxes. Advective temperature displacements as large as 170¿ occur only when average permeability is 10-15 m2 and can be detected using isotopic or petrologic thermometry. These temperature deviations reflect local permeability heterogeneities and could be useful in identifying high-permeability paleochannels. These results also imply, however, that temperature profiles taken from only one portion of a heterogeneous contact aureole may not be regionally representative or used to infer the overall advective-conductive thermal structure of the large-scale hydrothermal system. ¿ 1998 American Geophysical Union