Stressful environments are often said to increase the expression of inbreeding depression. Alternatively, Crow's “opportunity for selection” (the squared phenotypic coefficient of variation) sets a limit to how much selection can occur, constraining the magnitude of inbreeding depression. To test these hypotheses, we planted self- and cross-fertilized seeds of Brassica rapa into a factorial experiment that varied plant density and saline watering stresses. We then repeated the experiment, reducing the salt concentration. We observed considerable inbreeding depression, particularly for survival in the first experiment and growth in the second. Both stresses independently depressed plant performance. Families differed in their amounts of inbreeding depression and reaction norms across environments. Outcrossed progeny were sometimes more variable. Stresses had small and inconsistent effects on inbreeding depression and, when significant, tended to diminish it. Levels of phenotypic variability often predicted whether inbreeding depression would increase or decrease across environments and were particularly effective in predicting which traits display the most inbreeding depression. Thus, we find little support for the stress hypothesis and mixed support for the phenotypic variability hypothesis. Variable levels of phenotypic variation provide a parsimonious explanation for shifts in inbreeding depression that should be tested before invoking more complex hypotheses.