Stray voltage (usually <10 V) can occur in farms. However, very little information is available related to sheep. In addition, little work has been carried out on the effects of the contextual conditions under which the animals are submitted to stray voltage. The aims of this study were (i) to determine the threshold voltage at which lambs start to express avoidance behaviour and (ii) to test if the contextual conditions (i.e. choice v. no-choice conditions) influence the determination of the threshold voltage inducing avoidance behaviour. Six-month-old female lambs fed ad libitum were trained to eat palatable pellets from one or two metallic feeders situated at the end of a 4-m long raceway. Voltage was then applied during a 2-min test to either the only feeder available (no-choice test, 1F, n = 13) or to the first of the two feeders in which the lamb started to eat (choice test, 2F, n = 13). The 1F lambs had to stop eating to avoid the voltage, whereas the 2F lambs were allowed to switch to the non-electrified feeder to carry on eating without any stray voltage. Stray voltage was applied every day, in steps of 0.5 V (AC, 50 Hz), from 0 up to 8 V. For voltages higher than 4.5 V, 2F lambs spent less time eating and ate less in the electrified feeder compared with the non-electrified feeder, and their latency to switch to the non-electrified feeder was shorter. In addition, a transient modification of behaviour was observed at 1.5 V. For 1F lambs, a decrease in the quantity of feed eaten was found for voltages higher than 5 V, although the time spent eating in the electrified feeder was not modified. Finally, 1F lambs urinated more during or just after the 2 min test than 2F lambs for voltages above 5 V. Although lambs with no choice experienced stray voltage as a negative event (increased occurrence of urination), they carried on eating in the electrified feeder whatever the voltage. Therefore, the contextual conditions in which animals are exposed to stray voltage influence their subsequent reactions: the first clear behavioural reaction threshold is easier to detect in choice than in no-choice conditions.