Previous reviews of stress, and the stress hormone cortisol, in fish have focussed on physiology, due to interest in impacts on aquaculture production. Here, we discuss cortisol in relation to fish welfare. Cortisol is a readily measured component of the primary (neuroendocrine) stress response and is relevant to fish welfare as it affects physiological and brain functions and modifies behaviour. However, we argue that cortisol has little value if welfare is viewed purely from a functional (or behavioural) perspective-the cortisol response itself is a natural, adaptive response and is not predictive of coping as downstream impacts on function and behaviour are dose-, time- and context-dependent and not predictable. Nevertheless, we argue that welfare should be considered in terms of mental health and feelings, and that stress in relation to welfare should be viewed as psychological, rather than physiological. We contend that cortisol can be used (with caution) as a tractable indicator of how fish perceive (and feel about) their environment, psychological stress and feelings in fish. Cortisol responses are directly triggered by the brain and fish studies do indicate cortisol responses to psychological stressors, i.e., those with no direct physicochemical action. We discuss the practicalities of using cortisol to ask the fish themselves how they feel about husbandry practices and the culture environment. Single time point measurements of cortisol are of little value in assessing the stress level of fish as studies need to account for diurnal and seasonal variations, and environmental and genetic factors. Areas in need of greater clarity for the use of cortisol as an indicator of fish feelings are the separation of (physiological) stress from (psychological) distress, the separation of chronic stress from acclimation, and the interactions between feelings, cortisol, mood and behaviour.