The number of bacterial cells in soil that form colonies on nutrient agar represent a small fraction of the direct microscopic counts (DMC). The colony-forming cells have larger cell dimensions than the very small ("dwarf") cells which represent the majority of the DMC. This may indicate that the dwarf cells are species unable to form visible colonies on agar, or that they swell to normal dimensions when growing. Indigenous bacterial cells were separated from soil by density gradient centrifugation and fractionated according to diameter by filtration through polycarbonate filters. Each filtrate was studied with respect to DMC, cell dimensions, colony-forming cells (visible colonies and microcolonies), and cell dimensions during growth on the agar. The calculated average percent viability was only 0.2% for cells with diameters below 0.4μm, about 10% for cells with diameters between 0.4 and 0.6μm, and 30-40% for cells with diameters above 0.6μm. Only 10-20% of the viable cells with diameters <0.4μm increased their diameter to >0.4μm prior to growth. Thus, size change during starvation and growth cycles did not explain the high numbers of dwarf cells observed by microscopy. The results show that despite the relatively low number of colony-forming bacteria in soil, the species that form colonies may be fairly representative for the medium size and large cells, which constitute a major part of the bacterial biovolume. Thus plate counting could be a useful method to count and isolate the bacteria accounting for much of the biovolume in soil. The origin of the dwarf cells is still unclear, but the low number of small cells that increased in size seems to indicate that the majority of these bacterial cells are not small forms of ordinary sized bacteria.