In the Camargue in southern France, catches of tabanids in a Manitoba trap in 1976–78 gave seasonal, daily and hourly records of numbers. Concurrently, some interactions of the flies with a group of free-ranging horses were observed directly. The tabanids were present from May to October annually, with the main peak of abundance between late May and early July. The number trapped in each hour of the day increased with the ambient temperature in the morning, was stable during the hottest time of the day and declined in the late afternoon. Their alighting on a horse always elicited reactions aimed at dislodging them. When tabanids became very abundant and active, the horses would move as a group from where they were feeding to a traditional open bare area (‘chomadou’) where they would stay without feeding, sometimes for hours, even though large numbers of tabanids were still worrying them. Analyses of data on an index of the abundance of tabanids on or about the horses in a group showed the horses did not reduce the fly load by moving to and resting on their chomadou area. The number of responses by a horse to the tabanids on or about it was shown to be dependent not only on the number of flies but on their activity. Analysis of the responses revealed a negative effect of wind speed on the activity component which was separate from its known negative effect on the number of flies present. A hypothesis is proposed that movement to the chomadou can be explained as the horses going to an area ‘known’ to be usually exposed to the prevailing wind.