A new species of soil-dwelling entomopathogenic nematode Heterorhabditis hepialus killed up to 100% (mean=72%) of root-boring caterpillars of a ghost moth Hepialus californicus in coastal shrub lands. When unchecked, ghost moth caterpillars killed bush lupine, Lupinus arboreus. Here we describe this strange food chain. Although unappreciated by ecologists, entomopathogenic nematodes are widespread and probably one of the most important groups of natural enemies for underground insects. The free-living infective juvenile (IJ) of entomopathogenic nematodes searches for host insects in the soil. A single IJ can kill a host, although several often invade together. After entering the host through a spiracle or other orifice, the IJ regurgitates its symbiotic bacterium, Photorhabdus luminescens, which kills the host within 48 h. The bacteria digest the cadaver and provide food for the exponentially growing nematode population inside. The bacteria produce antibiotics and other noxious substances that protect the host cadaver from other microbes in the soil. When the cadaver is exhausted of resources, IJs break the host integument and can disperse. As many as 420,000 IJs can be produced within a large ghost moth caterpillar. Surface soil of the lupine rhizosphere is the primary habitat of IJs of H. hepialus. Attracted to waste gases emitted by insects, the 0.5-mm-long IJs can move 6 cm/day through moist soil. Prevalences of H. hepialus ranged from as high as 78% of rhizospheres in some lupine stands to almost zero in others, but it was absent from no stand at our study site. Field intensities ranged from 0.003 IJs/cm3 of soil to 7.5 IJs/cm3, and correlated roughly with prevalences among sites. Few ghost moth caterpillars (mean=6.7) succeeded in entering lupine roots where prevalence of H. hepialus was highest, and this stand had lowest mortality (0.02) of mature bush lupine. In the three stands with lowest prevalence (mean = 2%) of this nematode, many caterpillars (mean = 38.5) entered roots, and lupine mortality was high (range = 0.41–1.0). Old aerial photographs indicate that the stands with highest recent nematode prevalence have had little or no mass die-off of lupine over the past 40 years. The photos depict repeated die-offs of lupine during the past four decades in stands with lowest recent prevalence of the nematode. This pattern leads us to entertain the hypothesis that the nematode affects vegetation dynamics indirectly through a trophic cascade. Dispersal of entomopathogenic nematodes is little understood. We found that air drying of soil extirpates H. hepialus and speculate that this nematode is dispersed during the wet season in moist soil bits on the exterior of fossorial insects and mammals. H. hepialus colonized some previously unoccupied lupine rhizospheres during the wet winter-spring season and, obversely, became extinct from some rhizosperes as soil dried in summer. Root-feeding insects have only recently been recognized as a force in communities, and the regulation of these important herbivores is still largely an ecological terra incognita. All evidence indicates that entomopathogenic nematodes are found throughout terrestril ecosystems, and we propose that trophic chains similar to those described in this report should not be uncommon.