VTM (Vegetation Type Map) plots comprise a huge data set on vegetation composition for many parts of California collected mostly between 1929 and 1935. Historical changes in vegetation have been inferred by sampling these areas many decades later and evaluating the changes in plant dominance. VTM plots can not be precisely relocated, and it has been assumed that errors resulting from this problem are inconsequential or can be eliminated by comparison with a composite of multiple contemporary plots. This study examines that assumption for southern California shrubland landscapes by comparing the differences in species composition between closely positioned VTM-sized plots. Comparing shrub species density in 400-m² plots separated by 30 m (center to center), I found that all species exhibited considerable differences in density even over this short distance. This patchiness in shrub distribution could lead to major errors in historical reconstructions from VTM plot data. Two methods are proposed for dealing with this problem. One is to collect multiple samples from the vicinity of the VTM plot and use the observed spatial variation to set bounds on the temporal changes required to represent significant historical change. The other is to look at broad landscape changes reflected in the averages observed in a large sampling of sites.