Experiments investigating mortality in the soil seedbank are aided by using only seeds that are initially viable and capable of remaining ungerminated (hereafter “persistent seeds”). However, seed mortality experiments often use heterogeneous populations containing persistent, nonviable, and germinable individuals. In this investigation we developed and compared nondestructive tests for isolating persistent seeds of two weed species characterized by physical seed dormancy (dormancy imposed by a water-impermeable seed coat): velvetleaf and ivyleaf morningglory. Individual seeds were weighed, steeped in water (hereafter “steepate”) for 48 h, and then assayed for imbibition. These seeds were then subjected to persistence assays conducted under controlled conditions (60 d in hydrated soil under 25/15 C day/night temperatures, 14-h photoperiod). Persistent seeds were less likely to imbibe and more likely to produce steepates with low electrical conductivity compared with germinable and nonviable seeds. For velvetleaf, persistent seeds were best segregated by comparing changes in steepate conductivity during 4 to 48 h of soaking, with the corresponding classification and regression tree (CART) model making few false discoveries (false discovery rate for persistence; FDRp = 8.6%, n = 93) and many true positive classifications (true positive rate for persistence; TPRp = 100%, n = 85). For ivyleaf morningglory, both a change in steepate conductivity from 4 to 48 h of soaking and imbibition status after soaking accurately separated persistent seeds (accuracy measures of corresponding CART models: FDRp = 5.6%, n = 150; TPRp = 100%, n = 142). Thus, for species with physical seed dormancy, we recommend use of steepate conductivity and imbibition status after soaking for isolation of persistent seeds. These seeds can then be used to optimize experiments on mortality in the soil seedbank. Nondestructive tests for isolating persistent seeds of species characterized by physiological seed dormancy require further research.