By scoring the chromosome number of developing embryos, we show that the sex ratio bias of the African social spider Stegodyphus dumicola Pocock is the result of an overproduction of female embryos. Only 17% of 585 embryos sexed from 14 egg sacs were male, a significant departure from a 1:1 sex ratio. We also explored the possibility of direct control of the sex of individual offspring in this species by examining the variance in the number of males per sac and the spatial distribution of male and female embryos within the sacs. We postulated that a variance in the number of males per sac lower than binomial (i.e., underdispersed or precise sex ratios) or a non-random distribution of male embryos within the sacs would suggest direct control of the sex of individual offspring. We found that the variance in the number of males per sac was indistinguishable from binomial and significantly larger than expected under exact ratios. Likewise, the spatial distribution of male embryos within three sacs examined was no more clustered than expected by chance. The sex ratio biasing mechanism in this species, therefore, apparently only allows control of the mean sex ratio but not of its variance. We present randomization and Monte Carlo methods that can be applied to test for departures from a random spatial arrangement of male and female embryos in an egg mass and for departures from binomial or exact ratios when not all members of a clutch have been sexed.