The halter-restraint test consisted of fitting each sheep with a halter and tethering it to a post. The trainer sat quietly next to the ewe. The amount of time each sheep spent actively pulling on the halter in a 2 min period was recorded, as well as occurrences of vocalizations. During shearing, sheep were restrained in a manner typical of shearing (first on their hindquarters then on their sides) and had a bladeless pair of shears passed over their bodies for 2 min. The number of kicks and vocalizations were recorded. To determine if differences existed among the experimental groups prior to training, one-way analyses of variance (ANOVA) were conducted for all de- pendent variables on the pre-test measures alone. For these analyses, the in- dependent variable was handling condition (gentle, restrained, and control). For all variables showing significant condition effects before training, three- way repeated measures analyses of covariance (ANCOVA) were performed, with pre-test scores serving as the covariate. However, for those dependent variables showing no significant differences prior to handling, three-way ANOVAs for unequal subjects were conducted (source×handling condi- tion Xtest session) using post- and re-test scores. The purpose of this test was to determine what differences, if any, existed between the groups after han- dling, and whether these differences were retained for 3 weeks. Newman- Keuls post-hoc tests were used to determine specific significant differences between means for all analyses, and a was set at P< 0.05. The decoy sheep participated in all three of the test sessions, but her data were not included in any of the analyses.