We present the first meta-analysis of a key measure in conservation biology: minimum viable population (MVP) size. Our analysis is based on studies published since the early 1970s, and covers 141 sources and 212 species (after filtering 529 sources and 2202 species). By implementing a unique standardization procedure to make reported MVPs comparable, we were able to derive a cross-species frequency distribution of MVP with a median of 4169 individuals (95% CI = 3577–5129). This standardized database provides a reference set of MVPs from which conservation practitioners can generalize the range expected for particular species (or surrogate taxa) of concern when demographic information is lacking. We provide a synthesis of MVP-related research over the past 30 years, and test for ‘rules of thumb’ relating MVP to extinction vulnerability using well-known threat correlates such as body mass and range decline. We find little support for any plausible ecological and life history predictors of MVP, even though correlates explain >50% of the variation in IUCN threat status. We conclude that a species’ or population’s MVP is context-specific, and there are no simple short-cuts to its derivation. However, our findings are consistent with biological theory and MVPs derived from abundance time series in that the MVP for most species will exceed a few thousand individuals.