Assessment of mental state is an important goal in animal welfare. The mental states of animals have been assessed using judgement bias but not in livestock. There is also the possibility that emotional arousal (emotional reactivity) could be indicative of mental state. This study aimed to induce differences in the mental states of sheep and then determine if this altered their judgement biases and emotional reactivity. To test for judgement biases, twenty-six ewes were trained to respond to two different bucket positions in a 3x6 m pen. One position was positively reinforced with food so the sheep approached it quickly; the other was associated with a fan-forced blower so the sheep avoided it. Following training, half the sheep were subjected to unpredictable, aversive events (2-3 per day for 3 weeks) that might modify mental state while control sheep were maintained in a standard environment. Events included restraint, mixing with unfamiliar sheep, exposure to a dog, simulated shearing, ear tattooing, foot bathing and transportation. Treated and control sheep were then tested for judgement biases by presenting the bucket in ambiguous locations between the learned positions, and responses were recorded. Following judgement bias testing, emotional reactivity was tested by measuring behavioural responses to events that were novel, sudden or unexpected. Judgement bias data were analysed by GLM and emotional reactivity behaviours by ANOVA. The unpredictable treatment induced negatively skewed judgement biases with these sheep approaching the ambiguous buckets less frequently than the controls (approach frequency: 38% and 46% respectively, S.E=2.38%, P=0.024). There were no treatment differences in emotional reactivity, with all sheep displaying similar behavioural responses to events (P>0.05). The occurrence of judgment biases suggests that exposure to unpredictable aversive events affected the mental state of sheep. This methodology may be useful in providing another component for assessing welfare states.