The "burglar alarm" theory of bioluminescence was investigated by determining predation rates of a nocturnal teleost predator, Porichthys notatus, on nonluminescent kelp mysids illuminated by dinoflagellate flashes, between the fall and spring of 1989/1990. Mysids (Holmesimysis costata) were placed in aquaria containing varying concentrations (0 to 40 cells/ml) of the dinoflagellate Pyrocystis fusiformis and a single midshipman fish. Controls used P. fusiformis during their luminescence-inhibited day phase. Mysid swimming movements readily stimulated dinoflagellate luminescence. Flashes and prey strikes were observed simultaneously by image-intensifying and infrared video cameras on a splitscreen monitor. Predation rates increased at dinoflagellate concentrations of 3 to 15 cells/ml and decreased below controls at levels>20 cells/ml. Videotape analysis showed that at low concentrations (2 to 5 cells/ml), strike success rates exceeded 75% if prey were previously illuminated by a flash, but dropped below 50% at higher cell densities. Increased predation was attributed to luminescence revealing prey position. The decrease at higher concentrations was considered to be due to greater flash frequency providing a more diffuse and confusing target. The study demonstrates the effects of secondary luminescence on zooplankton predation at normally encountered dinoflagellate concentrations.