Heat flow measurements along much of the San Andreas fault (SAF) constrain the apparent coefficient of friction (??app) of the fault to <0.2, much lower than laboratory-derived friction values for most geologic materials. However, heat flow data are sparse near the creeping section of the SAF, a frictional "asperity" where the fault slips almost exclusively by aseismic creep. We test the hypothesis that the creeping section has a substantially higher or lower ?? app than adjacent sections of the SAF. We use numerical models to explore the effects of faults with spatially and temporally heterogeneous frictional strength on the spatial distribution of surface heat flow. Heat flow from finite length asperities is uniformly lower than predicted by assuming an infinitely long fault. Over geologic time, lateral offset from strike-slip faulting produces heat flow patterns that are asymmetric across the fault and along strike. We explore a range of asperity sizes, slip rates, and displacement histories for comparing predicted spatial patterns of heat flow with existing measurements. Models with ??app ??? 0.1 fit the data best. For most scenarios, heat flow anomalies from a frictional asperity with ??app > 0.2 should be detectable even with the sparse existing observations, implying that ??app for the creeping section is as low as the surrounding SAF. Because the creeping section does not slip in large earthquakes, the mechanism controlling its weakness is not related to dynamic processes resulting from high slip rate earthquake ruptures. Copyright 2006 by the American Geophysical Union.