Summary Population viability analysis (PVA) is widely used for assessing the extinction risk faced by endangered species and for evaluating the effects of potential management strategies. However, most PVAs are done in urgency and with limited data. It is therefore important to know how the amount and quality of available ecological data affect the predictions of PVA. How can PVA results be interpreted when we have reliable, long-term population data? When a species is endangered on islands, as are many endemic bird species, what are the likely causes of the decline, and what is the probable impact of new threats? This paper investigates these issues by modelling 26 years of monitoring data on the population of Capricorn silvereyes Zosterops lateralis chlorocephala Campbell & White of Heron Island on the Great Barrier Reef. Australia. A PVA with parameters estimated from a relatively short (5-year) period of data collection produced very different predictions of extinction risk compared with a PVA based on more comprehensive ecological data spanning 15–25 years. This indicates that caution should be exercised when using PVA as an indicator of absolute extinction risk. Despite remaining viable over the past 26 years, the Heron Island silvereye population is predicted to be at risk of extinction within the next 100 years, with the magnitude of risk depending on the severity of impact of a range of potentially threatening factors. Increased mortality due to the introduction of new predators or diseases, inbreeding depression or frequent severe storms greatly increased the extinction risk. Habitat reduction, if not very severe, is unlikely to have a large impact. Similar threats and impacts may apply to other island bird populations. Movement or dispersal between neighbouring islands had a mitigating effect on extinction probability and allowed extinct island populations to be re-established. However, very small islands could only support sink populations... [ABSTRACT FROM AUTHOR]