Virtual fencing presents a promising solution for grazing management, providing flexibility, individual monitoring, potential improvements in land management, and enhanced animal welfare. However, welfare issues related to electric stimuli and species-specific learning abilities need consideration. This study aimed to investigate if sheep could learn and be fenced by a fully automated virtual fencing system. Eight Norwegian White ewes with twin lambs, all naïve to virtual fencing, were randomly selected at the beginning of September 2018. Fitted with Nofence virtual fence collars designed for goats, the ewes were subjected to daily behavioural observations over nine days. The Nofence technology is a fully automated system that uses GPS coordinates, that trigger an audio cue when an animal crosses the boundary. If the animal does not turn, it receives an electric stimulus. The study was divided into three periods: an adaptation period (2 days), period 1 (5 days), where the first virtual boundary was introduced on day 1, and period 2 (4 days) when a second boundary was introduced. Additionally, the ewes wore heart rate monitors for at least one hour daily. Throughout the study, the ewes received a total of 218 audio cues and 38 electric stimuli, indicating an approximate individual mean of 27 audio cues and 5 electric stimuli received The number of audio cues and electric stimuli significantly decreased over days in both experimental periods. In period 2, the number of audio cues was significantly higher compared to period 1, while the success ratio in avoiding electric stimuli remained stable across days. The experimental period significantly influenced the frequency of standing/walking behaviour, being more common in the adaptation period compared to periods 1 and 2. However, the period did not have a significant effect on the percentages of lying or grazing behaviour. During heart rate monitoring, four ewes received one electric stimulus each, resulting in an immediate heart rate increase that returned to baseline within minutes. The present study demonstrates that sheep wearing a fully automated system quickly learned to associate audio cues with electric stimuli. The ability to generalise experiences from one virtual boundary to another emphasises their learning ability. Minor changes in activity were likely influenced by factors other than the virtual fencing system. The short-lasting increase in heart rate as a response to electric stimuli suggests a temporary stress reaction. Overall, this study provides valuable insights into the potential of virtual fencing for sheep management.