Simple SummaryGlobally, many unwanted dogs enter rescue shelters. Shelter staff often avail of behavioural tests as an early screening tool to identify areas of concern to minimise the welfare risk associated with long-term kennelling and failed adoptions. A number of requirements need to be verified in order for a test to become a useful assessment tool, including how reliable and accurate the measurements are. For these tools to be widely used, they need to be feasible and reproducible. We refined a previously validated temperament test for shelter dogs’ assessment, developed in Italy, and applied it to two populations of shelter dogs in the UK. The test measured dog behaviour in the kennel, sociability towards people and other dogs, docility to leash, playfulness, cognitive skills, and reactivity. The test proved easy to replicate, with key outcomes that are consistent with existing research on this topic. Furthermore, an additional experiment provided support for the use of fake dogs instead of real ones to assess sociability to dogs. However, we also highlight the importance of interpreting these data with caution, and advocate the use of behavioural tests as a partial screening tool to be used in conjunction with more extensive behavioural and welfare monitoring.AbstractThis study assessed the feasibility and reproducibility of a previously validated temperament test (TT) for shelter dogs. The test was developed to measure dog behaviour in the kennel, and traits of sociability towards people and other dogs, docility to leash, playfulness, cognitive skills, and reactivity. We introduced the use of differently sized fake dogs to check their appropriateness in correctly assessing sociability to dogs to broaden its applicability (as the original study used real stimulus dogs). We hypothesised that dogs’ responses may be modulated by the body size of the stimulus dog presented. The reduction analysis of the TT scores extracted five main dimensions (explaining 70.8% of variance), with high internal consistency (alpha > 0.65) and being broadly consistent with existing research. Behavioural components that were extracted from the fake dog experiment showed that dogs are likely to show signs of anxiety and fear toward both the real and fake dog. Dogs’ responses towards a real vs. fake stimulus were significantly correlated (p < 0.05) and they were not affected by the size of the stimulus (p > 0.05). We discuss the importance of interpreting these data with caution and use behavioural tests as a partial screening tool to be used in conjunction with more extensive behavioural and welfare monitoring.