Can measures of human–animal relationships track changes in commercial farms in the way animals are handled? Hemsworth et al. (2002) conducted a large scale intervention study based on a special training program to improve the attitudes of stockpeople to dairy cows. Following the training program, the stockpeople were found to use a lower number of aversive handling treatments and a higher number of gentle or positive handling treatments. The effects were substantial: the number of positive behaviours used on the intervention farms was more than double that on control farms, while the number of forceful negative behaviours was a fifth. The flight distance of the cows was found to be significantly lower on the farms on which the stockpeople had followed the training programs, however, the effect was small: flight distance on control farms was 4.49 m while on intervention farms it was 4.16 m. Furthermore, there were no significant effects on the occurrence of flinch, step and kick responses during milking. Thus, the cows’ behaviour was a poor predictor of changes in the way that the stockperson handled the cows. A similar study with stockmen on a large commercial piggery (Coleman et al., 2000) found that an appropriate course aimed at targeting attitudes to animals did result in a substantial reduction in use of aversive handling by stockpeople who followed the course. However, there was no significant reduction in the pigs’ tendency to withdraw when the stockman approached, although there was a non-significant trend in this direction (P = 0.11). These two intervention studies show that it would be very difficult to track even quite substantial improvements in the stockmen’s handling of the animals by using changes in the animals’ responses to people. These results support our concern about the validity of using these tests in on-farm welfare assessment or in animal welfare audits.
Do different measures measure the same thing?
If tests are valid, then two tests that claim to measure the same thing should produce the same results. Few comparisons between alternative measures of human–animal relations have been done. Most data available is on the comparison between approach distance and flight distance. Interestingly, approach distance and flight distance appear not to be alternative ways of measuring the same thing, but rather measures of different aspects of the relationship between people and animals. In beef cattle, Murphey et al. (1981) found that flight distance and approach distance were not correlated across different breeds and herds of cattle. The two different ‘‘distances’’ may also reflect different aspects of how the animals are treated by the stockperson: Hemsworth et al. (2000) found that flight distances of dairy cows tended to be negatively correlated with the extent that the stockperson used positive interactions, while approach distances tended to be correlated with the use of negative handling by the stockperson. Experimental treatments also appear to affect flight and approach distances differently. Jago et al. (1999) found that hand feeding of calves reduced approach distances but had no effect on flight distance. Although, Krohn et al. (2001) reported that extra handling affected both, the approach distance appeared to be more sensitive to the age at which the handling was received than was flight distance. Clearly, approach distance and flight distance reflect different aspects of the relationship between people and animals, but we have insufficient information, as well as a lack of conceptual models, to determine just what these different aspects may be. Thus, we have insufficient data to judge how other measures of human–animal relationships are interrelated.
Effects of context
To be valid, the tests should not be strongly affected by minor ‘‘contextual’’ variables that do not strongly impact the animals’ degree of fear. There are many different ways of doing these tests, and some of the differences may affect how the animals react to the people. For example, different studies may use different size enclosures, allow different time latencies for the animals to make contact, either the person or the animal can enter the enclosure first (the latter making it difficult to control for initial distances when the person entered) and the person can be sitting, standing, or even on a horse. These may seem relatively minor points, and if standardised within an experiment, will not necessarily result in any confounding of the results. However, we know little about how these extraneous variables do affect our measures, and they will limit our ability to evaluate these factors and identify a valid on-farm test. Furthermore it may not be possible to control for these factors on farms, making a standardised test more difficult to develop.