Introduction

Sheep, like most large herbivores, live in groups with some form of social organization, in which the distances maintained between members of a group may reflect relationships between individuals, as well as the cohesion of the group as a whole. From the measurement of nearest neighbor distances, evidence for associations between pairs of animals has been found in a diversity of domestic breeds of sheep [1]. We present a statistical method which can also detect individual differences in general "sociability", or the tendency to be close to others.

Data collection

Four groups of 10 female Scottish Blackface sheep grazed in separate 0.4 ha plots. Each group of sheep was observed for a 2-h period on 16 days spread over a 12-week period. During these 2-h observation periods, the identity of each sheep's nearest neighbor was recorded at either 5 or 10 minute intervals.

Statistical analysis

For each group of sheep 16 matrices were constructed. In each matrix the rows contained the proportion of observations in which each sheep was observed to be every other sheep's nearest neighbor on one of the observation days. A single matrix was then obtained by summing the 16 matrices and the relative frequencies with which individual sheep were any other sheep's nearest neighbor obtained by calculating the column totals. The data was then examined by comparing the variance of the single summed matrix and the variance of the column totals, with the variances of the same values derived from 1000 simulations. In each simulation for any day sheep numbers were randomized, but the nearest neighbor counts remained as observed on each day. This process allowed us to test whether differences between individuals in the frequency with which they were nearest neighbors were consistent across days. The calculations required for the matrix analysis were programmed using the statistical package Genstat [2].

Results

The single summed matrix and column totals for group 1 is shown in Table 1. Similar tables were constructed for the other groups. Within groups 1, 3 and 4 there were significant differences between individuals in the frequency with which they were any other sheep's nearest neighbor (P<0.001). In group 1, for example, sheep 1 had a much higher value than sheep 4 in the column totals (Table 1), indicating a greater degree of sociability. Associations could be identified between 4 pairs of sheep in groups 1, 2 and 4 (P<0.001) and between 3 pairs of sheep in group 3 (P<0.01).

Table 1. Matrix for group 1, showing summed proportions of observations in which each sheep was the nearest neighbor (columns) of every other sheep (rows), with column totals. Values considered to identify associated pairs are indicated thus*.

Conclusions

This method can detect differences in the frequency with which individuals are found close to others in a group, in addition to identifying associations between pairs. It may therefore provide a useful tool for describing the social behaviour of animals in groups.

References

1. Arnold, G.W. (1985). Association and social behaviour. In: A.F. Fraser (ed.). Ethology of Farm Animals, pp. 233-246. Amsterdam: Elsevier.
2. Genstat 5 Committee (1993). Genstat 5 Release 3 Reference Manual. Oxford: Clarendon Press.