We test the hypothesis that the origin of the diverse suite of aftershock mechanisms following the 1989 M 7.1 Loma Prieta, California, earthquake is related to the post-main-shock static stress field. We use a 3-D boundary-element algorithm to calculate static stresses, combined with a Coulomb failure criterion to calculate conjugate failure planes at aftershock locations. The post-main-shock static stress field is taken as the sum of a pre-existing stress field and changes in stress due to the heterogeneous slip across the Loma Prieta rupture plane. The background stress field is assumed to be either a simple shear parallel to the regional trend of the San Andreas fault or approximately fault-normal compression. A suite of synthetic aftershock mechanisms from the conjugate failure planes is generated and quantitatively compared (allowing for uncertainties in both mechanism parameters and earthquake locations) to well-constrained mechanisms reported in the US Geological Survey Northern California Seismic Network catalogue. We also compare calculated rakes with those observed by resolving the calculated stress tensor onto observed focal mechanism nodal planes, assuming either plane to be a likely rupture plane. Various permutations of the assumed background stress field, frictional coefficients of aftershock fault planes, methods of comparisons, etc. explain between 52 and 92 per cent of the aftershock mechanisms. We can explain a similar proportion of mechanisms however by comparing a randomly reordered catalogue with the various suites of synthetic aftershocks. The inability to duplicate aftershock mechanisms reliably on a one-to-one basis is probably a function of the combined uncertainties in models of main-shock slip distribution, the background stress field, and aftershock locations. In particular we show theoretically that any specific main-shock slip distribution and a reasonable background stress field are able to generate a highly variable suite of failure planes such that quite different aftershock mechanisms may be expected to occur within a kilometre or less of each other. This scale of variability is less than the probable location error of aftershock earthquakes in the Loma Prieta region. We successfully duplicate a measure of the variability in the mechanisms of the entire suite of aftershocks. If static stress changes are responsible for the generation of aftershock mechanisms, we are able to place quantitative constraints on the level of stress that must have existed in the upper crust prior to the Loma Prieta rupture. This stress level appears to be too low to generate the average slip across the main-shock rupture plane. Possible reasons for this result range from incorrect initial assumptions of homogeneity in the background stress field, friction and fault geometry to driving stresses that arise from deeper in the crust or upper mantle. Alternatively, aftershock focal mechanisms may be determined by processes other than, or in addition to, static stress changes, such as pore-pressure changes or dynamic stresses.