Predator responses to gradients in prey density have important implications for population regulation and are a potential structuring force for subtidal marine communities, particularly on rocky reefs where herbivorous sea urchins can drive community state shifts. On rocky reefs in southern California where predatory sea otters have been extirpated, top‐down control of sea urchins by alternative predators has been hypothesized but rarely tested experimentally. In laboratory feeding assays, predatory spiny lobsters (Panulirus interruptus) demonstrated a saturating functional response to urchin prey, whereby urchin proportional mortality was inversely density‐dependent. In field experiments on rocky reefs near San Diego, California, predators (primarily the labrid fish California sheephead, Semicossyphus pulcher) inflicted highly variable mortality on purple urchin (Strongylocentrotus purpuratus) prey across all density levels. However, at low to moderate densities commonly observed within kelp forests, purple urchin mortality increased to a peak at a density of ~11 urchins/m2. Above that level, at densities typical of urchin barrens, purple urchin mortality was density‐independent. When larger red urchins (Mesocentrotus franciscanus) were offered to predators simultaneously with purple urchins, mortality was density‐independent. Underwater videography revealed a positive relationship between purple urchin density and both the number and richness of fish predators, but these correlations were not observed when red urchins were present. Our results demonstrate highly variable mortality rates across prey densities in this system and suggest that top‐down control of urchins can occur only under limited circumstances. Our findings provide insight into the dynamics of alternate community states observed on rocky reefs.