OBJECTIVE: To assess the reliability of a French version of the Horse Grimace Scale (HGSfv). STUDY DESIGN: Prospective, randomized, clinical study. ANIMALS: The operated (OP) group included 13 horses undergoing elective surgery. The positive (PC) and negative control (NC) groups included seven colicking horses and eight exercising sport horses, respectively. METHODS: Photographs were extracted from videos of the horses' heads. Videos were taken before and immediately after surgery in OP, on arrival of the horse in PC, and at rest in their stalls in NC. Pictures were evaluated by three anaesthetists [Diplomates (DIPs)] and four riders (RIDs) using Horse Grimace Scale translated into French (HGSfv) at two points, 2 weeks apart (E1 and E2). Each evaluator gave each image a score (1-3) for six identified facial action units. The scores given by DIPs and RIDs were compared using a Wilcoxon test. Intra- and inter-evaluator reliability were assessed using Spearman correlation tests (r(s)) and intra-class coefficients (ICCs), respectively. RESULTS: RIDs and DIPs gave significantly higher scores in the PC group than in the NC group [RIDsE1PC 5.0 (4.2-9.8) versus RIDsE1NC 2.2 (0.0-6.5), p = 0.02; RIDsE2PC 5.2 (3.2-9.5) versus RIDsE2NC 2.0 (0.2-5.8), p \textless 0.01; DIPsE1PC 4.0 (1.3-6.3) versus DIPsE1NC 2.2 (1.0-4.7), p = 0.04; DIPsE2PC 2.7 (1.0-6.0) versus DIPsE2NC 1.0 (0.0-2.3), p = 0.03]. Scores given by RID or DIPs 2 weeks apart were highly correlated [r(s) (RIDsE1, RIDsE2) r = 0.86, p \textless 0.0001] and [r(s) (DIPsE1, DIPsE2) r = 0.81 p \textless 0.0001]. The ICC between RIDs and DIPs in E1 and E2 was 0.94 (0.92-0.95) and 0.91 (0.89-0.93), respectively. The specificity and sensitivity of the HGSfv was 94% and 43%, respectively. CONCLUSIONS AND CLINICAL RELEVANCE: Using the HGSfv, knowledge of horses rather than specialization in veterinary anaesthesia and analgesia appears to differentiate horses with visceral pain from those assumed to be pain free.