It has been reported that grooming behaviour may be related to status (calving, stress), health (mastitis) and production (milk yield). Consequently, accurate long-term monitoring of grooming behaviour could advantageously serve as a PLF solution to give alarms or well-being indicators. The main activities of dairy cows (resting, moving, eating, and so on) can be monitored indoors 24/7 using a real-time location system (RTLS). During the EU-PLF project (2012-16), the CowView RTLS (GEA) was assessed on 160 cows, equipped with tags, in our freestall barn enhanced with mechanical swinging brushes (Delaval). From the large positioning dataset collected (1 location/animal/s), we extracted grooming behaviour (scratching using the brush) with good sensitivity (80%) and a suitable positive predictive value (60%). The accuracy of our grooming detection algorithm was evaluated by comparing the results output against video analyses. Then, 23 dairy cows were monitored for 32 days, and patterns of grooming activity were analysed to study variations between cows and between days. 24-h video recording was re-visualized to explain the false detections observed, i.e. a cow detected in the brush area without grooming or a cow classified as grooming but not localized in the area. The kinetics of daily time spent grooming showed high day-today and inter-individual variations. Noisy tags appeared to be responsible for most of the false detections. Once smoothed, the kinetics modelled offer promising applications for detecting alterations in cow grooming behaviour patterns.