Validation of qualitative behaviour assessment for dairy cows at pasture

Abstract

Qualitative Behaviour Assessment (QBA) has been validated as a measure of emotional state in dairy cows kept indoors. We investigated the validity of QBA as a measure of emotional state of cows at pasture. We focused on intra- and inter-observer reliability and the selectivity of QBA. First, 5 observers watched 20 videos of dairy cows at pasture 4 times (resulting in 400 observations). The observers performed QBA using the fixed list of 20 terms proposed in Welfare Quality. The first axis of the principal component analysis performed on these observations represented emotional valence (PC1, from irritable to content) and a second axis represented arousal (PC2, from calm to active). Kendall's concordance coefficients (W) and intraclass correlation coefficients (ICC) within and between observers were calculated for PC1 and PC2. Intra-observer reliability was mainly high (KW > 0.75 and ICC > 0.75), except for 3 observers during the first session for PC1, for whom reliability was moderate. Inter-observer reliability was low to moderate for PC1 and moderate to high for PC2. Second, two observers performed live QBA on a herd of cows at pasture during 3 contexts designed to induce different emotional valences and arousal levels: AM, in the morning after milking (when cows are active); PM, in the afternoon before milking (when cows are mainly resting); and during handling to collect cows at pasture for afternoon milking (when cows may be stressed). The effect of context on PC1 and PC2 was investigated using linear mixed effects models. AM and PM contexts had higher scores on PC1 than handling context, indicating a more positive emotional state during AM and PM than during handling. PM context had lower scores on PC2 than AM and handling contexts. Thus, QBA at pasture was able to discriminate between contexts that should differ in emotional valence and arousal. Thirdly, an observer assessed the behavioural responses of cows to handling by direct observations followed by QBA. Pearson correlation coefficients were calculated between the number of behavioural responses and the coordinates on PC1 and PC2. The more cows were observed trotting, the more the cow's mood was perceived as negative (r = -0.71) and the more cows were observed galloping and turning, the more the cow was perceived as excited (r = 0.77 and 0.60). In conclusion, the QBA appears to be a valid measure of the emotional and arousal state of dairy cows at pasture, but inter-observer reliability could be improved.