Can we measure human–animal interactions in on-farm animal welfare assessment? Some unresolved issues : Validity
Validity is much harder to assess than reliability, especially in animal welfare or behavioural research, because we usually lack a well-defined standard with which we can compare our test. Validity of measures of animals’ responses to people can be judged by experimentally altering the animals’ degree of fear of people, and assessing to what extent the measures are able to detect this. However, often validity is judged circumstantially by the extent that the test is not affected strongly by other variables, and by the extent that different types of tests produce the same conclusion. Validity can be reduced if the measure is sensitive to changes in supposedly minor parameters of the test that should not influence the results. For example, on-farm measures of fearfulness of people should not be strongly affected by the clothing the people wear, or the location of the test. Finding strong effects of such variables would throw doubt on the validity of the test. In order to judge the validity of a measure, we need to be clear about what exactly the measure is supposed to be measuring. For example, are we measuring the animal’s responses to people in general or to one person (the farmer) in particular? Are we assessing the animal’s general responses to people, or a particular type of response, e.g. degree of fearfulness, ease of handling, etc.? Arewe interested in using the animals’ responses to people to assess howthey are handled, or as one more indicator of their welfare? Unfortunately, in many studies, the precise goal of the measure is sometimes not made clear.The most convincing evidence that the use of aversive handling causes animals to become fearful of people, and that this is apparent in the way the animals react to people, comes from experimental studies in which animals are deliberately handled gently or roughly. Numerous studies on swine, cattle and poultry (reviewed in Hemsworth and Coleman, 1998; Hemsworth, 2003) show that aversive handling of animals will lead them to avoid people while gentle handling will lead them to approach people. Furthermore, the avoidance of people by aversively handled animals is often associated with physiological indicators of stress, such as elevated corticosteroids. Most recently, for example, Breuer et al. (2003) handled some heifers aversively and others gently. Clear effects of the treatment were noted: aversively handled heifers showed a greater flight distance, increased approach latencies, fewer interactions with the handler, were more agitated in a crush and had higher cortisol concentrations in the presence of the handler. The relationship is even more convincing where animals have been trained to avoid one person through aversive handling and approach another person through gentle handling. For example, Rushen et al. (1999a) trained cows to recognise two different people, one of whom always treated the cows well and the other always treated the cows roughly. After several treatments, the cows approached the gentle handler closer than the aversive handler. That this indicates some fearfulness of the people was shown by the fact that residual milk and heart rate were higher when the cows were milked in the presence of the aversive handler. This study provides clear evidence that the reaction of an animal to people can predict how the animal was handled and indicate whether or not the animal is likely to show any physiological signs of stress. However, the relationship between the animals’ responses to people and other stress responses can be variable. A subsequent study (Munksgaard et al., 2001), which effectively used the same similar experimental protocol as Rushen et al. (1999a), but less aversive handling, found the same magnitude of effect upon the animals’ withdrawal responses from the people but found no effects upon residual milk. Thus, even in highly controlled experiments, the relationship between the animals’ behavioural responses to people and other stress measures can be variable and observations of animals’ responses to people may not predict their physiological stress responses. Furthermore, the results of the studies show the difficulties in establishing a standardised ‘‘cut-off’’ point that could be used in welfare audits to determine that the animals’ degree of fear of people is sufficiently high to warrant some intervention. Many studies have reported a statistically significant difference between the effect of aversive handling and gentle handling on animals’ responses to people. However, a highly significant difference between handling treatments does not mean that the effect of the treatment is large, only that it is highly unlikely to have occurred by chance. When we look at the size of some of the effects of type of handling on animals’ responses to people, we discover that these are often quite small. For example, in the study by Breuer et al. (2003) the magnitude of the differences cast some doubts on whether aversively handled heifers were substantially more fearful of people than gently handled ones. Even though approach latencies were higher for aversively handled heifers than for gently handled ones, the aversively handled heifers still took an average of only 165 s to approach within 1 m of the experimenter, compared to only 120 s for gently handled ones. Thus, the magnitude of the results does not give the impression of highly fearful animals, nor of a large difference between gently and aversively handled animals. We must ask the question whether these differences are large enough to be useful in differentiating between farms where the animals are handled well and those where there is a handling problem that is sufficiently large to reduce the animals’ welfare.