Experiment 1: Validity of the aversion race

. treatment pen , and the amount of force required to move the animal down the race Ž . summed over the four sections of the race were recorded. 2.3. Statistical analysis For all three experiments, the total amount of time and the summed force taken to move through the race as well as the latency to enter the race were analysed by analyses Ž . of variance using the general linear model procedure of SAS 1990 . The model consisted of two factors: the treatments applied to the animal, which varied between Ž . experiments experiment 1, df s 3; experiment 2, df s 4; experiment 3, df s 2 and the Ž trial number, which was treated as a repeated measures factor experiment 1, df s 1,10; . experiments 2 and 3, df s 3,6 . To correct for initial differences between cows in moving down the race before treatments were applied, we included the measure on the Ž . first trial as a covariate df s 1 . We tested for interactions between trial number and the treatment. Contrast statements were used to compare treatments within each experiment Ž . each contrast df s 1 . Due to large variation, data were log transformed before analyses, but untransformed values are presented for clarity. In all experiments, prelimi- nary analyses demonstrated no effect of week, group or any interactions between these factors and treatment, and these factors were removed from the final analysis.

3. Experiment 1: Validity of the aversion race

The purpose of this experiment was to test the empirical validity of the aversion race by determining how well it differentiates between treatments known to differ in aversiveness. Previous studies have demonstrated that providing food decreases the distance cows keep from people, while hitting animals increases the distance cows keep from the people, clearly demonstrating that dairy cattle find food to be positive and Ž . hitting to be negative Munksgaard et al., 1997; Rushen et al., 1999 . Non-aggressive forms of tactile contact such as brushing or petting are also reported to decrease fear and Ž . make animals easier to handle Boissy and Bouissou, 1988; Boivin et al., 1992 . 3.1. Methods Fifty-four lactating multiparous cows were randomly allocated to one of four treatments. Ž . a. Control n s 14 . The person stood still, facing the cow with hands in pockets and did not talk or handle the animal. Ž . b. Brushing n s 14 . The cows were brushed on the neck with a dairy cattle grooming brush while the person spoke softly to them. Ž . c. Food n s 13 . The cow was simultaneously offered a handful of hay and molasses flavoured feed. Ž . d. Hitrshout n s 13 . The person hit the animal forcefully four times on the head or neck with an open hand and shouted at the cow for the total duration of the treatment. Ž All treatment periods were 2 min in duration. Two groups of four animals one . animal per treatment were tested each week. Cows passed through the race three times Ž a day over 4 days for a total of 12 trials i.e., passage through the corridor and . treatment . Three different handlers were used. Each cow was always treated by one person only and each person did all four treatments for an equal number of cows. Order of treatments during the day was balanced across groups. 3.2. Results Overall, the latency for cows to enter the race decreased over the course of the Ž . experiment P - 0.05, Fig. 2a . There was no significant treatment effect on average Ž . Ž . Ž . Fig. 2. Behaviour of cows meanSE during experiment 1. a Latency to enter the race, b time to move Ž . through the race, and c force required to move through the race. latency and no trial by treatment interaction, although there was a non-significant tendency for cows on the hitrshout treatment to take longer to enter the race than cows Ž . on other treatments P - 0.10. . On trial 6, cows on the hitrshout treatment took Ž . significantly longer to enter the race than control cows P - 0.05 and cows on the food Ž . treatment P - 0.01 . On trial 8, cows on the hitrshout treatment took longer to enter Ž . the race than cows on each of the other three treatments P - 0.05 . There was a large decrease from trial 1 to trial 2 in the amount of time to move down Ž . the race Fig. 2b for cows on all treatments. Overall, the time to walk through the race Ž . differed significantly between treatments P - 0.001 . Cows on the hitrshout treatment Ž . took longer to walk through the race than cows on the other three treatments P - 0.01 . Cows given food took less time to walk through the race than control cows on trial 4 Ž . P - 0.05 . Brushed cows took longer to walk through the race than cows given food on Ž . Ž . trials 4 and 5 P - 0.05 and from control cows on trial 4 P - 0.05 . The amount of force required to move the cow through the race reflected the amount of time required to walk through the race; there was a significant treatment effect Ž . Ž . P - 0.001 but no trial by treatment interaction Fig. 2c . On average, cows on the hitrshout treatment required more force to move through the race than cows on the Ž . other three treatments P - 0.01 , while cows given food required less force to move Ž . through the maze than cows on the brush treatment P - 0.05 . Control cows required Ž more force to move through the maze than cows given food on trials 11 and 12 both . comparison P - 0.05 .

4. Experiment 2: Methods for moving cows