Sexual reproduction in flowering plants involves double fertilization, the union of two sperm from pollen with two sex cells in the female embryo sac. Modern plant breeders increasingly seek to circumvent this process to produce doubled haploid individuals, which derive from the chromosome-doubled cells of the haploid gametophyte. Doubled haploid production fixes recombinant haploid genomes in inbred lines, shaving years off the breeding process(1). Costly, genotype-dependent tissue culture methods are used in many crops(2), while seed-based in vivo doubled haploid systems are rare in nature(3) and difficult to manage in breeding programmes(4). The multi-billion-dollar maize hybrid seed business, however, is supported by industrial doubled haploid pipelines using intraspecific crosses to in vivo haploid inducer males derived from Stock 6, first reported in 1959 (ref. 5), followed by colchicine treatment. Despite decades of use, the mode of action remains controversial(6-10). Here we establish, through fine mapping, genome sequencing, genetic complementation, and gene editing, that haploid induction in maize (Zea mays) is triggered by a frame-shift mutation in MATRILINEAL (MTL), a pollen-specific phospholipase, and that novel edits in MTL lead to a 6.7% haploid induction rate (the percentage of haploid progeny versus total progeny). Wildtype MTL protein localizes exclusively to sperm cytoplasm, and pollen RNA-sequence profiling identifies a suite of pollen-specific genes overexpressed during haploid induction, some of which may mediate the formation of haploid seed(11-15). These findings highlight the importance of male gamete cytoplasmic components to reproductive success and male genome transmittance. Given the conservation of MTL in the cereals, this discovery may enable development of in vivo haploid induction systems to accelerate breeding in crop plants.