Testing judgement biases of animals may provide insight into their affective states; however important questions about methodologies need to be answered. This experiment investigated the effect of repeated testing using unreinforced, ambiguous cues on the response of sheep to a go/no-go judgement bias test. Fifteen sheep were trained to differentiate between two locations, reinforced respectively with feed (positive) or with the presentation of a dog (negative). The responses to nine ambiguous locations, positioned between the positively and negatively reinforced locations, were tested repeatedly over 3 weeks. Sheep exhibited a symmetrical gradation in response to ambiguous locations between the positive and negative reinforcers. There was a significant decline (P = 0.001) in the total number of approaches to the ambiguous positions over time (weeks). This effect of time suggests that sheep learnt that the ambiguous locations were unrewarded. This result supplies evidence of a limitation identified in current judgement bias methodology, due to repeated testing, which has the potential to provide misleading results.