Urination in transponder-controlled roughage feeding stations is a widespread undesirable behaviour of group-housed horses. Urination on hard surfaces, such as the floor of the stations, is contrary to the natural elimination behaviour of horses because they prefer to urinate on soft, absorbent surfaces, and it increases ammonia emissions around the stations. The following aspects were analysed: a) urination as potential displacement activity during feed anticipation, b) absence of appropriate elimination areas in the stable and c) ammonia odour as a trigger stimulus. We observed a group of 33 horses in three different situations: 1) baseline situation, 2) provision of elimination areas containing an absorbent substrate in front of the feeding stations and 3) neutralisation of ammonia odour in the feeding stations. In the baseline situation all horses were observed, regardless whether they urinated in the feeding station or not. In the other two situations, only the urinating-horses were observed. We analysed 5 h of video per day, recorded on 4 days from seven feeding stations in each situation. We used an information theory approach, calculating three different (generalized) linear mixed effects models and all according sub-models. In the baseline situation, the horses showed that the horses urinated often after ‘ground exploration’, and there was more ground exploration in the urinating-horses than in the non-urinating-horses. In addition, a higher percentage of the mares than of the geldings urinated during at least one station visit, and mares urinated more often per visit than geldings. Before urination, the horses in most cases lowered the head toward the hay container to trigger the sensor and cause the station to open the partition and make the hay accessible. The time span between lowering the head and access to feed could be perceived as too long by the horses (maximum duration: 30 s) and lead to urination as a displacement activity. In addition, urination never occurred when the hay was accessible, only when the hay was inaccessible, closed, opening or closing. This leads us to conclude that urination is related to feed anticipation. The frequency of urination bouts was not reduced by installing additional elimination areas or neutralising the urine odour, so additional elimination areas and urine odour do not seem to have a role in the undesirable behaviour. Further research is needed to investigate a displacement activity or a classical conditioning in more detail to prevent urination by horses in automated feeding stations.