Abstract: We describe a method for quantifying relative sociability within a group of animals, which is defined as the tendency to be close to others within the group and based on the identification of nearest neighbours. The method is suitable for groups of animals in which all individuals are visible and identifiable and has application as a tool in other areas of behavioural research. A sociability index (SI) is calculated, which is equivalent to the relative proportion of time that an individual spends as the nearest neighbour of other animals in the group and is scaled to have an expectation of 1.0 under the null hypothesis of random mixing. Associated pairs, which are animals seen as nearest neighbours more often than would be expected by chance, are also identified. The method tests for consistency across a number of independent observation periods, by comparison with values obtained from simulations in which animal identities are randomised between observation periods. An experiment is described in which 8 groups of 7 grazing sheep were each observed for a total of 10, one-hour periods and the identities and distances away of the 3 nearest neighbours of each focal animal recorded at 5-min intervals. Significant within-group differences in SIs were found in four of the groups (P < 0.001). SIs calculated using the nearest neighbour, two nearest neighbours or three nearest neighbours, were generally highly correlated within all groups, with little change in the ranking of animals. There were significant negative correlations between SIs and nearest neighbour distances in five of the groups. It was concluded that there was no advantage in recording more than one neighbour to calculate the SI. Advantages of the SI over other methods for measuring sociability and pair-wise associations are discussed.