Comparison of methods to assess fear of humans in sows

The aims of this study were to compare three different methods to assess fear of humans in sows, and identify the most practical and reliable method for on-farm welfare assessment. Fifty-five crossbred sows (Landrace × Large White × Duroc) of different parities and stages of pregnancy were each subjected to three tests—Test O: response of free moving sows to a stationary observer in an open field test; Test S: response of sows in stalls to approaching hand contact; Test G: response of free moving sows in groups to approach by an observer. Each test was repeated after a period of no more than one week and sows behavioural responses were scored according to pre-determined scales of approach–withdrawal behaviour. Although previously validated in another study, Test O showed poor repeatability (rs = 0.153, P > 0.05), whereas repeatability was good for both Test S (rs = 0.777, P < 0.001) and Test G (rs = 0.628, P < 0.001). A Kendall coefficient of concordance suggested that there was no overall significant relationship among the three tests (W = 0.507, P > 0.05). Rank correlations showed a weak but significant relationship between Test O and Test S (rs = 0.270, P < 0.05), a stronger correlation between Test S and Test G (rs = 0.404, P < 0.01), but no significant correlation between Test O and Test G (rs = 0.197, P > 0.05). There were no significant effects of parity or stage of pregnancy on fear responses in any of the three tests. Since they are practical, repeatable and measure the same level of underlying fear of humans in sows, we recommend Test S and Test G for on-farm welfare assessment in different housing systems.