Although a major component of fitness, male reproductive success is generally extremely difficult to estimate. As a result, genetic methods and maximum likelihood models have been developed to estimate male parentage, but all are limited in practice by the degree of genetic variation observable. Scoring individuals phenotypically at a large number of random loci exhibiting dominance (e.g. RAPD markers) may provide a means of detecting sufficient genetic variation. Dominance, however, represents a loss of information and therefore greater variation in the estimate of paternity. A mixture model describing mating in a population is presented to quantify the trade-off between marker types when estimates of male fertility are sought. A sample size 1.5-2.0 times greater is required for dominant markers under some conditions to obtain the same confidence in fertility estimates as for codominant markers, although with large sample sizes the fertility estimates are similar for either marker type. Since the number of dominant DN A markers is not limited in the same manner as is the number of codominant protein markers, one's confidence in the estimates can be increased above that possible from proteins by surveying additional loci. However, for a fixed sample size a trade-off exists between the number of progeny assayed per female and the number of loci surveyed. In many cases more progeny per female provide better estimates of fertility than more loci.