Abstract
To ensure good animal welfare in laboratory research and in stockbreeding severity ratings of the animals´ wellbeing are essential. The current study investigated how valid raters can evaluate different severity degrees of clinical appearance and how ratings might be influenced by factors other than the severity itself. Ninety-seven people rated the severity degree (none, mild, moderate, or severe) of the clinical appearance of mice seen in eight different images. The images also differed in the perspective in which they had been taken (entire mouse or head only). The raters differed with regard to their experience of working with laboratory animals and were subsequently divided into three groups-beginners, advanced, professionals. Generalisability theory was applied to examine the contribution of the different rater (raters themselves and experience) and image facets (actual degree of severity and perspective) to the overall data variability. The images showing the extreme severity degrees were rated more homogenously and more precisely than were the images showing the intermediate degrees, as compared to the reference scores. The largest source of variance was the actual degree of severity, accounting for 56.6% of the total variance. Considering only the images showing the extreme severity degrees, this percentage rose to 91.6%, accounting almost exclusively for the found variance. In considering only the intermediate severity degrees, the actual degree of severity did not contribute to variance at all. The remaining variance was due to the raters and the interactions between raters, the actual degree of severity and the perspective. The experience of the raters did not account for any variance. Training in the assessment of severity degrees seems necessary to enhance detection of the intermediate degrees of severity, especially when images are used. In addition, good training material should be developed and evaluated to optimise teaching and to minimise wrong assessments.