In a mixed-mating population, the fitness of selfed individuals, relative to outcrossed individuals, can be estimated from observed changes of the inbreeding coefficient F. In general, three measurements of F, or two measurements of F and one of the selfing rate, spanning two generations are needed. If however, adult F is assumed constant among generations, the adult F and selfing rate of one generation are sufficient to estimate relative fitness. Estimates of the relative fitness of selfed individuals for 14 Mimulus populations, assuming equilibrium of adult F, averaged 0.38, which is significantly lower than the 0.50 threshold needed to favor selfing genotypes. However, estimates for some populations showed large variance and over the 14 populations, relative fitness did not correlate with selfing rate. Violations of the equilibrium assumption causes a positive bias of the estimate and a spurious negative correlation of estimates with selfing rate. Estimates of relative fitness based on two generations of F do not suffer these problems. The need for large sample sizes, multiallelic loci, and moderate levels of natural selfing limits the usefulness of using changes of F to infer inbreeding depression.