Theory predicts unified sex ratios for most organisms, yet biases may be engendered by selfish genetic elements such as endosymbionts that kill or feminize individuals with male genotypes. Although rare, feminization is established for Wolbachia-infected Eurema butterflies. This paradigm is presently confined to islands in the southern Japanese archipelago, where feminized phenotypes produce viable all-daughter broods. Here, we characterize sex bias for E. hecabe in continental Australia. Starting with 186 wild-caught females, we reared >6000 F1-F3 progeny in pedigree designs that incorporated selective antibiotic treatments. F1 generations expressed a consistent bias across 2 years and populations that was driven by an ~5% incidence of broods comprising ⩾80% daughters. Females from biased lineages continued to overproduce daughters over two generations of outcrossing to wild males. Treatment with antibiotics of differential strength influenced sex ratio only in biased lineages by inducing an equivalent incomplete degree of son overproduction. Brood sex ratios were nevertheless highly variable within lineages and across generations. Intriguingly, the cytogenetic signature of female karyotype was uniformly absent, even among phenotypic females in unbiased lineages. Molecular evidence supported the existence of a single Wolbachia strain at high prevalence, yet this was not clearly linked to brood sex bias. In sum, we establish an inherited, experimentally reversible tendency for incomplete offspring bias. Key features of our findings clearly depart from the Japanese feminization paradigm and highlight the potential for more subtle degrees of sex distortion in arthropods.