We investigate the relationship between frictional strength and clay mineralogy of natural fault gouge from a low-angle normal fault in Panamint Valley, California. Gouge samples were collected from the fault zone at five locations along a north--south transect of the range-bounding fault system, spanning a variety of bedrock lithologies. Samples were powdered and sheared in the double-direct shear configuration at room temperature and humidity. The coefficient of friction, 5, was measured at a range of normal stresses from 5 to 150 MPa for all samples. Our results reinforce the intuitive understanding that natural fault gouge zones are inherently heterogeneous. Samples from a single location exhibit dramatic differences in behavior, yet all three were collected within a meter of the fault core. For most of the samples, friction varies from 5 = 0.6 to 5 = 0.7, consistent with Byerlee's law. However, samples with greater than 50 wt % total clay content were much weaker (5 = 0.2--0.4). Expandable clay content of the samples ranged from 10 to 40 wt %. Frictional weakness did not correlate with expandable clays. Our results indicate that friction decreases with increasing total clay content, rather than with the abundance of expandable clays. The combination of field relations, analytical results, and friction measurements indicates a positive correlation between clay content, fabric intensity, and localization of strain in the fault core. A mechanism based upon foliations enveloping angular elements to reduce friction is suggested for weakening of fault gouge composed of mixed clay and granular material. We provide broad constraints of 1--5 km on the depth of gouge generation and the depth at which fault weakness initiates. We show that slip on the Panamint Valley fault and similar low-angle normal faults is mechanically feasible in the mid-upper crust if the strength of the fault is limited by weak, clay-rich fault gouge.