In the course of this study I examined 1849 soft-shelled turtles, including some incomplete alcoholic or dried specimens, such as those represented only by skulls or by other osteological material. Material was examined from each of the collections named below (except KKA), and these are mentioned in the text by the following abbreviations: AMNH American Museum of Natural History ANSP Academy of Natural Sciences, Philadelphia BCB Bryce C. Brown, private collection, Baylor University CM Carnegie Museum CNHM Chicago Natural History Museum INHS Illinois Natural History Survey, University of Illinois KKA Kraig K. Adler, private collection, data in letter dated January 8, 1960 KU Museum of Natural History, The University of Kansas LSU Louisiana State University MCZ Museum of Comparative Zoology, Harvard College MSU The Museum, Michigan State University NHB Naturhistorisches Museum Basel, Switzerland OU University of Oklahoma Museum, Division of Zoology SM Strecker Museum, Baylor University TCWC Texas Cooperative Wildlife Collection, Texas Agricultural and Mechanical College TNHC Texas Natural History Collection, The University of Texas TTC Texas Technological College TU Tulane University UA University of Alabama UI Museum of Natural History, The University of Illinois UMMZ Museum of Zoology, The University of Michigan USNM United States National Museum WEB William E. Brode, private collection, Mississippi Southern College WTN Wilfred T. Neill, private collection External measurements (listed under the section, "Variation") were taken by the writer by means of a Vernier caliper or a steel tape. Measurements of the skulls are in millimeters and tenths as taken by the writer with dial calipers. Partial wrinkling of the carapace at the edges of some specimens causes some error in measurements; consequently, length of plastron is used as the measurement of reference. Scattergrams based on external measurements were constructed. Some demonstrate considerable ontogenetic variation. An inspection of the scattergrams indicated regressions essentially linear in nature, but sometimes occasioned an arbitrary separation of samples into size groups to show ontogenetic variation; no secondary sexual differences could be discerned. Several ratios were developed from the measurements. The data correspond to the regression model 1A in "Statistical Methods" (Snedecor, 1956, sec. 6.13); consequently, the sample ratios indicate the slope of regression and are useful in comparisons. Sample-means and their estimated standard errors are compared graphically to show general trends in proportional characters. Comparisons of means and standard errors indicate statistical significance between populations if the sample-means plus or minus twice their standard errors do not overlap, but this method of comparison is valid only when comparing two samples (Pimentel, 1959:100).