Can suffering in non-human animals be studied scientifically? Apart from verbal reports of subjective feelings, which are uniquely human, I argue that it is possible to study the negative emotions we refer to as suffering by the same methods we use in ourselves. In particular, by asking animals what they find positively and negatively reinforcing (what they want and do not want), we can define positive and negative emotional states. Such emotional states may or may not be accompanied by subjective feelings but fortunately it is not necessary to solve the problem of consciousness to construct a scientific study of suffering and welfare. Improvements in animal welfare can be based on the answers to two questions: Q1: Will it improve animal health? and Q2: Will it give the animals something they want? This apparently simple formulation has the advantage of capturing what most people mean by ‘improving welfare’ and so halting a potentially dangerous split between scientific and non-scientific definitions of welfare. It can also be used to validate other controversial approaches to welfare such as naturalness, stereotypies, physiological and biochemical measures. Health and what animals want are thus not just two of many measures of welfare. They provide the definition of welfare against which others can be validated. They also tell us what research we have to do and how we can judge whether welfare of animals has been genuinely improved. What is important, however, is for this research to be done in situ so that it is directly applicable to the real world of farming, the sea or an animal’s wild habitat. It is here that ethology can make major contributions.