Protozoan determinations were made for a variety of samples from Ridley Creek, Pennsylvania, using Q-mode cluster analysis. Only a few species are found at a majority of sampling locations and most species were found in only a few. However, if one takes a series of samples, it is difficult to miss well-established species represented by a number of individuals. This was the judgment of Patrick et al. (1967), confirmed at that time by comparing results of different investigators sampling the same general site at different times during the same day. Results of Q-mode cluster analysis and principal coordinates analysis have shown that visually similar natural habitats are not more similar in terms of their protozoan communities than visually dissimilar habitats. Nor does spatial relationship appear to be a major factor. Specifically, samples from rocks in three riffles did not show greater within-riffle similarities than among riffles. Moreover, samples from eight different habitats showed no more tendency for faunal similarity within habitats than among habitats. Finally, among ten samples collected from the same rock, five of them from the same 7.5 x 7.5 cm area, protozoan distribution showed no readily interpretable pattern. Clearly, factors other than the apparent similarities of the habitats controlled the distribution of protozoans in the stream studied. Possibly, random invasion of habitats is a major factor in determining which of the array of species potentially capable of occupying a particular habitat at a particular time are actually present.