Reiterating arguments supporting the view that the ‘‘vast majority of mutations are deleterious,’’ Keightley and Lynch (2003) take issue with several aspects of the analysis and interpretation of a recent mutation-accumulation (M-A) experiment with Arabidopsis thaliana (Shaw et al. 2002). In hopes of eliminating misunderstanding, we here address their criticisms of this work. We begin, however, by considering prior evidence about two aspects of the distribution of mutational effects: the average effect on fitness of new mutations and the fraction of mutations that reduce fitness. Whereas the evolutionary implications of both aspects of distribution of mutational effects are profound, we note that evidence about the latter aspect is far weaker than that concerning the former. Direct evidence concerning directional tendency of effects of spontaneous mutations on fitness comes primarily from M-A studies, in which lines bearing mutations accumulated over numerous generations in the (near) absence of selection are grown contemporaneously with a control that is genetically similar to the founder from which the M-A lines were originally established. In such studies, the mean of fitness (in practice, a component of fitness), taken over all M-A lines, compared to the mean for the control provides evidence of directional tendency of the aggregate effects of new mutations, that is, the average of their effects. The best known M-A studies, those of Mukai (1964; Mukai et al. 1972) with Drosophila melanogaster, showed considerable reduction in fitness of late—generation M-A lines. Reexamination of these studies has called this evidence into question, in particular the validity of the control (Keightley 1996; Garcia-Dorado 1997; but see also Fry 2001). Accordingly, several more recent studies have focused on organisms for which it is possible to retain individuals in a quasi-inert state during the period of M-A, and later revive them for assays of fitness in experiments together with the advanced generation M-A lines. In studies of this kind, conclusive evidence of an overall tendency of mutations to reduce fitness, by the criterion that the mean of a fitness component averaged over M-A lines (mm-a) is significantly less than the control mean (m0) (i.e., mm-a , m0), has been obtained for fitness traits of several model organisms. These include Escherichia coli (population growth rate, Kibota and Lynch 1996), the nematode, Caenorhabditis elegans (productivity, survival to maturity, and generation rate, Vassilieva et al. 2000), Arabidopsis thaliana (number of seeds per fruit, Schultz et al. 1999), DNA repair-