Several studies on trait-mediated indirect interactions (TMIIs) have shown that predators can initiate trophic cascades by altering prey behaviour. Although it is well recognized that individual prey state alters antipredator and foraging behaviour, few studies explore whether this state-dependent prey behaviour can alter the strength of the ensuing tritrophic cascade. Here, we link state-dependent individual behaviour to community processes by experimentally testing whether hunger level and body size of prey altered antipredator behaviour and thus changed the strength of trophic cascades between predators and primary producers. In rocky intertidal tide pools on the California Coast, waterborne cues from the predatory seastar Leptasterias spp. (Stimpson) can cause the herbivorous snail Tegula (Chlorostoma) funebralis (A. Adams) to reduce grazing and flee tide pools, resulting in positive indirect effects on tide pool microalgae. However, we show that the strength of this behaviourally-mediated cascade may be contingent on prey hunger level and body size. During short field experiments at low tide, medium-sized snails that were either newly collected from the field or fed for 1 week in the laboratory mediated strong TMIIs because they grazed less when seastars were present. In contrast, no TMIIs occurred when medium-sized snails had been starved for 1 week because they continued grazing regardless of seastar presence. Newly collected small snails fled from seastars but did not mediate cascades because they ate little algae. Despite reaching an apparent size refuge from predation, many newly collected large snails fled from seastars, but those individuals that remained tended to graze the algae more quickly, resulting in unexpected negative indirect effects of seastars on algae cover. The implication of this pattern for the natural system is unclear. Because average hunger level and size of snails vary over time and space in nature, a mosaic of TMII strength may exist. Overall, the strength of tritrophic TMIIs in tide pools depended on individual prey state, supporting model predictions and adding to sparse empirical evidence. This outcome suggests that patterns occurring system-wide over the long term may be influenced by the state-dependent decisions made by the individuals present.