Two random samples of dolines in the Clarksville quadrangle, Tennessee (selected on the basis of the presence or absence of natural ponds), were statistically compared on the bases of long-axis orientation, vectorial mean long-axis orientation, agreement with systematic joint sets, and length of doline long axis versus doline mean width.
The two doline samples are not members of the same population of doline long-axis orientations, nor do the two samples have the same preferred long-axis orientations. Both samples have different vector mean long-axis orientations, and the vector mean long-axis orientations of dolines without natural ponds is statistically significant. The polymodal preferred doline long-axis orientations correlate with the three systematic joint sets in the study area, particularly the N20° to 40°E and N10° to 30°W sets.
Both doline samples belong to the same population of length/mean-width ratios and are related allometrically in the form L = 1.656 W1.024.