Repeatability of an attention bias test for sheep suggests variable influence of state and trait affect on behaviour

Understanding the effects of repeated testing on behaviour is essential for behavioural tests that are re-applied to the same individuals for research and welfare assessment purposes. Assessing the repeatability of behaviour can also help us understand the influence of persistent traits vs transient states on animal responses during testing. This study examined the repeatability of behavioural responses in an attention bias test developed for sheep as a measure of affective state. Sheep were assessed in the attention bias test three times (n = 81 sheep), with testing occurring at intervals of 1 year then 2 weeks. During testing, individual sheep were exposed to a dog located behind a window for 3 s in a 4 × 4 m arena, then the dog was obscured from view, removed and sheep behaviours were recorded for 180 s. We hypothesised that behaviours in the test would have moderate-high repeatability but that the mean behavioural responses would change over consecutive trials as sheep habituated to the test environment. To estimate repeatability, data were modelled using restricted maximum likelihood linear mixed-effects models, fitting animal ID as a random effect. Vigilance behaviour, defined as having the head at or above shoulder height, was moderately repeatable (r = 0.58). Latency to eat (r = 0.20) and duration spent looking towards the previous location of the dog (attention to the dog wall) (r = 0.08) had low repeatability. Mean latency to eat did not differ significantly between trials (P = 0.2) and mean vigilance behaviour tended to decrease over the trials (P = 0.07). Mean duration of attention to the dog wall significantly decreased across the trials (P