Extinction is a result of two key processes: ultimate drivers of extinction, which cause a species to decline to low numbers, and proximate drivers of extinction, which deliver the coup de grace to that small population. Islands are hot spots of avian extinctions but identifying the ultimate causes of insular avian extinctions is challenging when multiple agents of decline such as deforestation, human hunting, and predation by introduced vertebrates operate over extended timeframes. Here we combine Holocene palaeoecological data and historical observational data with probabilistic extinction-date estimators to illuminate the shifting drivers of decline for an endemic, flightless, New Zealand parrot species (ka over bar ka over bar po over bar , Strigops habroptila) as it approached near extinction in the late twentieth century. By comparing the prehuman (i.e., pre-1280 CE) distribution of ka over bar ka over bar po over bar with their historical (post-1769) distribution and overlaying this with data on prehuman and historical deforestation and feral dog pack observations, we show that the drivers of ka over bar ka over bar po over bar decline shifted in time across the North and South Islands of New Zealand. In the South Island, forest clearance appears to be the key correlate of local ka over bar ka over bar po over bar extirpation following Polynesian settlement. In the North Island the scarcity of historical (1769-current) ka over bar ka over bar po over bar observations, even in intact forest, suggests that other factors were operating and were potentially more intense (e.g., increased pressure from human hunting or higher rat densities). Historical-era ka over bar ka over bar po over bar observations did not generally overlap with observations of historical feral dog packs (predominantly comprising European dog breeds). Based on multiple extinction-date estimators, we estimate a 31-70-year lag between ka over bar ka over bar po over bar being lost in the South Island (predicted extirpation date range 1990-2006) compared to the North Island (1936-1959). Our results demonstrate that agents of decline differed between islands, and that this resulted in spatial variation in patterns of decline for ka over bar ka over bar po over bar .