Reliability of different behavioural tests for growing pigs on-farm

Behavioural tests might have the potential to evaluate certain aspects of welfare such as emotional state or the human-animal relationship. However, reliability assessments of these tests are rare. Therefore, in this study, different behavioural tests for growing pigs were assessed regarding their on-farm reliability. Eleven growing pig farms were visited three times each by two experienced observers. The farm visits took place with a time interval of one week in between the first and the second farm visits and a time interval of four weeks in between the first and the third farm visits, i.e. the second and the third farm visits were three weeks apart. During the second farm visit, exactly the same animals were tested while at the third farm visit, other animals were tested. The observers carried out a Novel object test (NOT), a Voluntary human approach test (VHAT), a Forced human approach test (FHAT) and a Human-Animal Relationship Test (HART) at group level in the home pen of the pigs. The results of the different farm visits and observers, respectively, were compared by a combination of statistical parameters: Spearman's rank correlation coefficient (RS), intraclass correlation coefficient (ICC), limits of agreement (LoA) and smallest detectable change (SDC). The interobserver reliability was in general acceptable for all tests (e.g. NOT: RS: 0.50, ICC: 0.56, SDC: 2.47, LoA: -2.49-2.28) to good (e.g. FHAT: RS: 0.98, ICC: 0.98, SDC: 0.23, LoA: - 0.22-0.25). The test-retest reliability, on the contrary, was rather insufficient: only in the comparison of the first to the second farm visits did RS and ICC reach values of >= 0.40 for the FHAT and HART (for some categories). However, the deviation between the two observers detected by SDC and LoA were above 10%. No reliability was detected for the other two comparisons as well as for all comparisons of the NOT and VHAT, (e.g. VHAT, first to second farm visit: RS: 0.15, ICC: 0.04, SDC: 30.2, LoA: - 24.5-38.6). Thus, it can be concluded that interobserver reliability was sufficient, while in the form of these behavioural tests used consistency over time at the moment remains insufficient. Hence, with regard to welfare assessment schemes, this aspect must be taken into account and future work should concentrate on recategorisation, enhancement of knowledge concerning the validity of these tests in different settings as well as possible alternatives. Nevertheless, the good interobserver reliability points out a potential usefulness of these tests for other purposes i.e. whether the animals need to be tested only once throughout their lives.