Qualitative behaviour assessment

Qualitative Behaviour Assessment (QBA) is a method that relies on the ability of human observers to integrate perceived details of behaviour, posture, and context into descriptions of an animal’s style of behaving, or ‘body language’, using descriptors such as ‘relaxed’, ‘tense’, ‘frustrated’ or ‘content’. Such terms have an expressive, emotional connotation, and provide information that is directly relevant to animal welfare and could be a useful addition to information obtained from quantitative indicators. Previous research with pigs, cattle, sheep and poultry consistently showed QBA to have high inter- and intra-observer reliability and to be coherent with quantitative behavioural and physiological measures, both when animals were assessed individually and at group level. Previous QBA work however was based on a Free-Choice-Profiling methodology that asks observers to develop their own descriptive terminologies, and is unsuitable for on-farm inspection. The aim of this study therefore was to design, and test the inter-observer reliability of, a fixed rating scale for QBA of cattle expression. This work was carried out with beef cattle, dairy cattle and veal calves. On the basis of previous QBA research with cattle and consultation with cattle experts, we designed a rating scale of 29 descriptors (32 for beef cattle). The rating scales were tested by three cohorts of four assessors, on 22 groups of veal calves and 22 groups of dairy cattle in Northern and Southern Italy, and on 21 groups of beef cattle in Southern Scotland. Assessors were given detailed instructions on the procedures of assessment and the use of the rating scale. The inter-observer reliability of the QBA scores attributed to the different cattle groups was tested using Kendall Correlation Coefficient W, and for beef cattle showed satisfactory reliability (W ≥ 0.70) for 20 out of 32 descriptors. For dairy cattle and calves this criterion was reached with only a few descriptors. However, comparison of Principal Component Analyses (PCA) of assessor scores within the three cattle groups showed remarkably similar emergent patterns of cattle expression, in which the first principal component (PC1) distinguished between positive and negative mood, and the second (PC2) differentiated these moods in low and high levels of arousal. These patterns were reproduced when descriptors with low loadings, low apparent welfare relevance, or with synonyms on the list, were removed from the assessor data sets (leaving 20 descriptors for each cattle group). For beef cattle, PC1 of the ‘reduced’ PCAs showed satisfactory inter-observer reliability (Kendall W=0.73; p<0.001) and explained 26-51% of the total variation between groups (depending on the assessor). PC1 scores also showed a high mean correlation (0.80; p<0.001) to the assessors’ scores of the descriptor ‘welfare overall’, which suggests that the distinction between positive and negative mood made by PC1 is directly relevant to beef cattle welfare. However, for dairy cattle and veal calves these emergent patterns, though present, were quantitatively weak. A subsequent video-based assessment of dairy cattle by 14 assessors using this rating scale found satisfactory reliability (Kendall W=0.73; p<0.001).We propose that PC1 may provide an integrative measure of positive and negative cattle emotion, to be accorded to single farm units through PCA of assessors’ scores on a 20-term rating scale. To calibrate the QBA measures of single farms, testing the QBA scale on a large sample of farm units is required to create a ‘benchmark’ data base. This in turn will allow identification of cut-off points on PC1 for unacceptable levels of negative mood/welfare. The application of QBA on farms is highly feasible and easy to learn. However, assessors must be experienced in observing cattle, and be given additional training in recognising cattle expression if required.