At what level of change in physiology or behaviour is an animal's welfare at risk? At first glance this question, addressed by Barnett and Hemsworth (1990), would seem to be of central importance to scientific studies of animal welfare. However, the question itself raises important issues which should be discussed. The notion of a point or 'cut-off line' at which welfare becomes 'at risk' necessarily implies that there is also an area in which welfare is 'not at risk'. The term 'at risk' is not specifically defined by Barnett and Hemsworth, but Barnett and Hutson ( 1987, p. 1 ) equate welfare which is at risk with unacceptable welfare. Therefore, on one side of this cut-off line welfare is acceptable, while on the other side it is unacceptable. Does the apparent scientific rigour of a cut-off line outweigh the dangers of mistakenly assessing the welfare of certain animals as being acceptable and not at risk? Consider three pigs kept in three different housing conditions, A, B and C. Measurement of their plasma free corticosteroid concentrations shows that the animals in Conditions A and B experience sustained rises of 45 and 30%, respectively, relative to the animal in Condition C (Fig. 1 ). Following Barnett and colleagues' ( 1987, 1990 ) suggestion that increases in free corticosteroid of > 40% represent a risk to a pig's welfare, we would conclude that only the welfare of the animal in Condit ion A is at risk. However, to many, the data would suggest that there is also a clear difference in the physiological responses of the other two animals that may be indicative of welfare problems in Condit ion B relative to Condit ion C. Should we, therefore, consider that both these pigs fall into the same 'not at risk' category, as Barnett and Hutson