The human-animal relationship is an essential part of farm animal welfare. Avoidance distance at the feeding rack (ADF) has been validated as a measure of human-animal relationship (HAR) for cows indoors, but no measure has been validated for cows at pasture yet. The aim of this study was to test the validity of 'avoidance distance at pasture' (ADP) as a measure of HAR for grazing dairy cows. We assessed the following validity criteria for both ADP and ADF: inter-observer reliability, test-retest reliability, and stability over the grazing season. We also assessed the selectivity of ADP by convergent validity with ADF. Two persons (randomly alternating between the experimenter carrying out the test and the observer watching the test) estimated ADF and ADP twice on 48 dairy cows. During the grazing season, one person measured both ADF and ADP on the 48 cows two days apart for test-retest reliability and 7 times (at 5-week intervals) for stability over the season. For these tests, the experimenter approached the cow from the front, with his hand held at approximately 45 • , and the distance between the experimenter's hand and the cow was measured at the moment of the cow's first avoidance movement. The inter-observer reliability, test-retest reliability, and stability over the season of ADF and ADP were analyzed at individual level using both reliability (e.g. Spearman's rank correlation (RS), Kendall's coefficient of concordance (KW)), and agreement parameters (e.g. smallest detectable change (SDC)). A linear mixed model was used to investigate the relationship between ADF and ADP. Inter-observer reliability was high (i.e. satisfactory) for ADF and ADP (RS>0.7). SDC was around 55 cm for ADF whereas it was 127 cm for ADP. Test-retest reliability was high for ADF (RS=0.7) and moderate for ADP (RS=0.5). Stability over the season was moderate for both ADF and ADP (0.5