The aim of this study was to investigate the effects of soft elastic stall partitions on the positions in own or neighbouring stalls by dairy cows, faeces and urine, and of contact between animal and partition. There were two designs of a soft elastic partition, here called I-partition and lambda-partition. The study was performed on a commercial dairy herd where the partitions had been developed. After clinical examination 16 lactating cows were selected and used as focal animals, and paired for video recordings. The animals were placed, according to their size, on stalls with different breadths and lengths. Each pair of focal animals was video recorded continuously for 9 days according to the following sequence: (1) with partition 3x24 h; (2) without partition 3x24 h; (3) with partition again 3x24 h. Before each such session each focal pair was again clinically examined. Behaviours were analysed per hour, summarised per 24h and analysed with a mixed linear model or Wilcoxon Rank Sum test. Without partitions more urine (p<0.05) and faeces (p<0.05) landed on the neighbouring cow's stall and the time standing 45 degrees to the manger was longer (p<0.01), whereas the time standing 90 degrees to the manger was shorter (p<0.01). Without partitions three out of the 16 cows were standing and three were lying parallel to the manger. Presence or absence of partitions did not influence the total lying time per 24 h. Cows with the lambda-partition defecated more seldom in neighbouring stalls than cows with the I-partition (p<0.01). It is concluded that without partitions between tied animals more eliminations on neighbouring stalls and more deviations from the 90 degrees position in relation to the manger occurred. This seems to be the first experimental study of how partitions influence the behaviour of tied cattle and the results support the general advice that tied cattle should have partitions.