Plants of the partially self-incompatible perennial herb Brassica cretica, derived from contralled cross- and self-pollinations within each of seven populations, were raised under uniform conditions and scored for two measures of developmental stability, flower asymmetry (quantified as the difference in length and width between opposite petals) and cotyledon asymmetry (quantified as the difference in the area of the two lobes of each cotyledon). The primary goals were to assess the level of heritable variation in asymmetry, the effect of selling on mean asymmetry, and the relationship between asymmetry and components of fitness. A paternal half-sibling analysis of data an flower asymmetry failed to detect significant levels of genetic variation at the within-population levee, whereas the between-population component reached significance for all measures of asymmetry. Analysis of family-structured data from another crossing experiment revealed significant between-population variation in cotyledon asymmetry and a tendency for inbred progeny to produce more asymmetric cotyledons than outbred progeny. However, the response to inbreeding was weak and differed in magnitude between populations. Judging from the ranking of populations, we found no support for the hypothesis that the mean expression of developmental stability is controlled by genomewide characteristics such as the level of inbreeding. Correlations between measures of asymmetry and fitness were too low to be declared statistically or biologically significant. The present study provides little evidence that flower and cotyledon asymmetry serve as more appropriate predictors of genetic health than conventional (direct) measures of fitness.