When biofilms (aggregations of bacteria and extracellular polymer secretions) in samples from the Carter Sandstone of Alabama were prepared for scanning electron microscopy (SEM) using different dehydration techniques, the organic material had visibly different textures and distributions. In order to assess whether the variation was attributable to sample preparation or to inherent biofilm heterogeneity, each of five techniques were tested 3 to 10 times on small (1 cm) pieces of the Carter Sandstone containing either a strain of bacteria cultured from and reintroduced into the rock, or an in situ biofilm grown by injection of nutrients through core samples. The techniques tested were (1) air drying alone; (2) fixation in 10% glutaraldehyde with air drying; (3) ethanol dehydration with hexamethyldisilazane (HMDS) drying [2.5% glutaraldehyde, ethanol dehydration, and HMDS]; (4) ethanol dehydration with critical-point drying; and (5) ethanol and acetone dehydration with critical-point drying. Unpreserved control samples were either imaged wet in an environmental scanning electron microscope (ESEM) or vacuum-dried for SEM. Observations were based on SEM microscopy of over 60 samples and study of over 150 photomicrographs. In both experiments, the original morphology of individual bacteria was best preserved by ethanol dehydration with HMDS drying, ethanol dehydration with critical-point drying, or ethanol-acetone dehydration with critical-point drying. Critical-point drying preserved bacteria but stripped away mucilaginous material, revealing filamentous structures within the biofilm. These filaments, along with masses of microspheres (nannobacteria?) and the smooth mucilaginous outer layer, also occur in wet samples studied by ESEM, and are, therefore, not dehydration artifacts. However, different sample preparation techniques accentuated different components of the heterogeneous biofilm, thus resulting in vastly different textures. The cultured bacteria produced a biofilm that had a different surface texture and was more susceptible to sample preparation artifacts than the in situ biofilm. Use of more than one sample preparation technique is recommended in order to avoid bias.