This paper takes a closer look at the subjectivity/objectivity relationship, as it plays a role in the science of animal welfare. It argues that subjective, experiential states in animals such as well-being and suffering are, contrary to what is often assumed, open to empirical observation and scientific assessment. The presumably purely private, inaccessible nature of such states is not an inherent property of these states, but derives from their misguided conception as ‘causal objects’ in mechanistic models of behaviour. This inevitably endows subjective experience with a ‘hidden’ status. However, subjective experience should be approached on its own conceptual grounds, i.e. as a perspective, in terms of ‘what-it-is-like-to-be’ a particular individual animal. Neither behaviour nor subjective experience then can be regarded as causal objects; they form an integrated, dynamic, expressive whole. The animal is perceived as an agent, whose perspective on a given situation is manifest in the way in which it interacts with and pays attention to that situation. In this framework, concepts of subjective experience such as enthusiasm, timidity, fear or contentedness, may be defined as categories of ‘attentional style’. Testing the scientific validity and reliability of such categories requires development of a qualitative methodology for the measurement of behaviour. Starting-points for such a method are put forward for discussion.