We describe a multilocus, marker-based regression method for inferring
interactions between genes controlling inbreeding depression in self-
fertile organisms. It is based upon selfing a parent heterozygous for
several unlinked codominant markers, then analyzing the fitness of pro
geny marker genotypes. If loci causing inbreeding depression are linke
d to marker loci, then viability selection is manifested by distorted
segregation of markers, and fecundity selection by dependence of the f
ecundity character upon the marker genotype. To characterize this sele
ction, fitness is regressed on the proportion of loci homozygous for m
arkers linked to deleterious alleles, and epistasis is detected by non
linearity of the regression. Alternatively, fitness can be regressed o
n the proportion of heterozygous loci. Other modes of selection can be
incorporated with a bivariate regression involving both homozygote an
d heterozygote marker genotypes. The advantage of this marker-based ap
proach is that ''purging'' is minimized and specific chromosomal segme
nts are identified; its disadvantage lies in low statistical power whe
n linkage is not strong and/or the linkage phase between marker and se
lected loci is uncertain. Using this method, in the wildflower Mimulus
guttatus, we found predominant multiplicative gene interaction determ
ining fecundity and some negative synergistic (nonmultiplicative) inte
raction for viability.