The affective state of an animal, which is thought to reflect its welfare, consists of both short-term emotional reactions and long-term general mood. Because this state is generated and processed by the brain, we used non-invasive measurement of such brain activity as a novel indicator variable and investigated the interplay of mood and short-term emotional reactions in animals. We developed a wireless sensor for functional near-infrared spectroscopy (fNIRS), which assesses cortical perfusion changes, and consequently neuronal activity. Mood differences were induced by barren and enriched housing in a total of nine sheep and we observed their brain reaction in response to the positive situation of being groomed. We detected a decrease in cerebral oxyhaemoglobin concentration ([O(2)Hb]) which persisted during grooming. The localisation of the decrease in the brain did not depend on the site where the stimulus was applied. Also, the intensity of the response did not depend on the intensity of the grooming stimulus and a sham stimulus did not evoke an [O(2)Hb] response as seen with a grooming stimulus. Thus, we conclude that the observed haemodynamic brain response was unlikely to reflect pure somato-sensory information. We then found that the amplitude of the [O(2)Hb] response was larger if sheep were in a supposedly more negative mood. This contradicts the common assumption that negative mood generally taints reactions to emotional stimuli. Our results also demonstrate the potential of fNIRS for assessing affective states in freely moving animals.