The idea of "indicator species" is a relatively old one with a huge associated literature. The idea that the existence or health of a single species somehow captures the state of an entire ecosystem, though, remains a challenge to those who, like us, study entire assemblages of invertebrates. This challenge is especially marked when the ecosystems of choice are rainforests. Nevertheless there remains a need for a measure of diversity somewhere between the single species and the entire assemblage, which, for invertebrates in rainforests, may add up to many hundreds of species. This need led one of us (RK), in the nineteen nineties, to formulate the idea of a 'predictor set': that is, a statistically defined set of species, from within a much larger assemblage, whose spatial or temporal distribution captures the patterns characteristic of the entire assemblage. This essay will trace the origin and development of this idea and present some current applications and future research questions. Of course, the use of single charismatic species as "flagship" or "umbrella" species is undoubtedly of conservation value and remains unchallenged. Key words: Predictor set, indicator species, flagship species, rainforest, moths