Altered assumptions about how different ecological factors limit population numbers may lead to different conclusions about the causes of decline and ultimate extinction of a small population. Here, alternative hypotheses for the local disappearance of the Serengeti plains study population of African wild dogs (Lycaon pictus) are examined in light of observations of density dependence, deterministic decline and frequent rapid fluctuations in population number. After a population crash from 60 individuals in 1975 to 30 individuals in 1976, the Serengeti plains population fluctuated around a mean value of 22 individuals for 16 years before local extinction occurred. Variation in population numbers was extreme, with inter-annual reductions in population size of at least 40% occurring three times. Several explanations are consistent with the observed trends in population size including outbreaks of various epizootics and competition with other predators. Monte Carlo simulation, with parameters set to reflect observed fluctuations, demonstrate that population extinction was likely from chance factors alone. In small and declining populations, for which precise data and controls are unavailable, determining the cause(s) of extinction usually will be impossible.