The objective of this study was to investigate whether dairy sheep, like cows, organise themselves into a well-defined sequence at entry into the milking parlour and to quantify the extent (i.e., proportion of flock members with a favourite position in the sequence) and consistency of the order over different periods of time (correspondence between two consecutive days and over periods of up to 13 weeks). A 51-head flock, comprising mainly British Milksheep, were studied under commercial husbandry conditions for 13 weeks and the sequence of entry into the milking parlour (type side-by-side, batch size 6) recorded, resulting in 18 readings of morning and 90 of afternoon milkings, respectively. The number of times an animal had occupied each position was plotted against the positions. The comparison of these frequency distributions of the 48 sheep studied for this purpose revealed that 75% of flock members grouped their choices around a favourite position in the milking sequence, resulting in a ± predictable milking order. The distribution patterns of the remaining 25% were less focused, but none of them scattered their choices evenly across the complete range of 48 possible places, demonstrating that, even among these, entry into the parlour was not at random. To test for short- and long-term consistency, analyses were carried out for morning and afternoon milking, over different periods of time, and over different groups of sheep, owing to the fact that the ewes joined the dairy flock successively. Correlations between the sequences of two consecutive days revealed, with r 2-values of up to 96%, very high degrees of consistency, all of which were significant at p<0.001. No significant differences in consistency between the groups could be detected. Correspondence tended to be greater between morning than between afternoon milkings (a.m.: r 2=73%, p.m.: r 2=65% on average, 0.05< p<0.1), which was probably due to different external conditions (e.g., practically no waiting time in collecting pen, no break, no audience, longer interval since previous milking). When testing for long-term consistency using concordance calculations, however, any significant correspondence disappeared ( W≤0.2298, p>0.05), leading to the conclusion that the milking order has to be regarded as a dynamic rather than a static phenomenon, being consistent only for periods of less than a week.