The relative importance of density-dependent and -independent processes in determining population density has been predicted to vary according to whether the population concerned is located near the centre or the periphery of the species' range. Thus, density-independent processes should be more pronounced near the periphery. The long-tailed wood mouse Apodemus sylvaticus in Iceland is at the northern and western edge of its geographical range. We estimated the autumn population density in an open habitat in south-western Iceland in 9 years out of 10 during 1996–2005 in order to monitor the annual maximum population size. Furthermore, we estimated population density and survival at c. 5-week intervals from September 2001 to October 2003 and from September 2004 to November 2005 in order to reveal the causes of variation in maximum population size. The estimated autumn population density was low, ranging from 2.7 to 8.9 mice ha−1 while spring densities ranged from 0.4 to 0.8 mice ha−1. Apparent monthly survival probabilities ranged from 0.4 to 0.7 per month in autumn and 0.7 to 0.9 in winter. Our results suggest that low temperature in early winter (October–December) is the major determinant of population density in the following autumn, explaining 74% of the variation in autumn population density. No significant correlation was found between either the NAO index or the NAO winter index and variation in wood mouse population density in autumn. Differential mortality in early winter results in variation in spring population size. This study shows clear evidence of density-independent control of a mammal population at the edge of its geographical range as opposed to the mostly density-dependent control previously recorded near its centre of distribution.