On-farm qualitative behaviour assessment of dairy goats in different housing conditions

This study reports the results of the first investigation on the use of Qualitative Behaviour Assessment (QBA) in dairy goats, using a fixed-list of descriptors specifically developed for this species. It aimed to verify whether QBA can be reliably used by observers with different backgrounds to differentiate between the emotional states of goats kept under different environmental conditions. Two trained observers simultaneously assessed 16 dairy goat farms (8 "Housed" (H) farms, where animals were observed in free stall pens, and 8 "Pasture" (P) farms, where animals were observed in open ranges), using a list of 16 QBA descriptors that were based on literature studies and discussed within a focus group of goat experts. One H farm was removed from analysis due to procedural error. The QBA scores were analysed together using Principal Component Analysis (PCA, correlation matrix, no rotation). Observer agreement for farm scores on PCA Components (PCs) and on separate QBA terms was investigated using Pearson and Spearman correlations respectively. The effects of housing system and observer on PC scores were analysed using analysis of variance (treatments = observer, housing system, and their interaction; block = farm). PCA identified three main components explaining 60.87% of the total variation between goat farms: PC1 (29.04%) ranged from "content/calm" to "frustrated/aggressive", suggesting a relationship to the animals' general mood; PC2 (19.70%) ranged from "curious/attentive" to "calm/bored", suggesting a relationship to the animals' level of arousal, and PC3 (12.13%) ranged from "sociable/playful" to "alert/agitated". The two observers showed a good level of agreement on the three PCA dimensions (PC1: r = 0.75, P = 0.001; PC2: r = 0.67, P = 0.006; PC3: r = 0.69, P = 0.004), and also on 7 out of 16 separate QBA descriptors (P. <. 0.05). Two additional descriptors showed a moderate level of agreement (P = 0.10). These results indicate an integrated PCA approach to QBA to be more robust. There were significant effects of housing system on both PC1 (ANOVA; P = 0.05) and PC2 (P = 0.02), indicating goats on P farms to be more "content/calm" and "curious/attentive" than goats on H farms. There was a significant observer effect on PC2 (P = 0.04), and a significant observer by housing interaction on PC3 (P = 0.009). In sum, these results suggest that QBA can be a reliable welfare indicator, used by observers with different backgrounds; however, further development of QBA training procedures is required to extend inter-observer reliability to all main expressive dimensions emerging from the analysis.