Inter-observer reliability of Qualitative Behavioural Assessments of sheep

Qualitative Behaviour Assessment (QBA) is whole-animal methodology that assesses the expressive qualities of animal demeanour using descriptors such as ‘relaxed’, ‘anxious’ or ‘content’. This study aimed to examine the inter-observer reliability of 12 fixed-list QBA descriptors for sheep that had been generated in a previous Free-Choice Profiling study by experienced animal welfare inspectors, based on the same video footage used in the current study. The 12 QBA terms were scored by two different assessor groups consisting of two veterinary science students and four veterinary surgeons (Group 1), and seven farm assurance inspectors (Group 2). The two assessor groups met and received training on different dates, and viewed the same 12 video clips of sheep showing a wide range of behavioural expressions in varying indoor and outdoor situations and housing systems. For each clip assessors scored the sheep expressions on each of the 12 QBA terms using a Visual Analogue Scale. Principal Component Analysis (covariance matrix, no rotation) was used to analyse assessor scores, both for Groups 1 and 2 separately, and together in one all-assessor group. All three analyses identified that over 70% of variance in sheep expression was described by two main components. For the all-assessor analysis the first Principal Component (PC1, 49% of variation) ranged from ‘content/relaxed/bright’ to ‘distressed/dejected/tense’, while PC2 (31%) ranged from ‘agitated/responsive/anxious’ to ‘dull/dejected/relaxed’. The 2 Principal Components identified by Groups 1 and 2 when analysed separately, were highly similar. The level of agreement between assessors in each group was evaluated using Kendall's coefficient of concordance (W). This produced W values of 0.83 (PC 1) and 0.84 (PC 2) for the all-assessor group, values of 0.90 (PC 1) and 0.86 (PC 2) for Group 1, and of 0.78 (PC 1) and 0.91 (PC 2) for Group 2. All values were significant at P<0.001. These results indicate that two trained assessor groups achieved high levels of inter-observer agreement using a list of 12 pre-fixed QBA terms developed for sheep. This study is the first to investigate QBA as a tool for assessing sheep behavioural expression, and its results support further exploration of the feasibility and validity of applying this method to the assessment of sheep welfare.