Two distinct approaches have emerged for the assessment of quality of life (QoL) and welfare in domestic dogs. One approach, which has so far been applied only to companion dogs, is derived from proxy assessment of QoL in human beings, with the owner or veterinarian acting as the proxy. Because dogs are a different species to human beings, assessment by proxy is even more challenging than when the subject being assessed is human. Our evaluation of published studies indicates that existing canine QoL instruments are imperfect, in part because of avoidable deficiencies such as failure to define QoL and using measures of health status as sole indicators of QoL. The second approach to QoL assessment, which stems from animal welfare science, is based upon objective measurement of behaviour and stress physiology, and has been applied mainly to dogs in laboratory and rescue kennels. We review these and our own recent studies, and conclude that although interpretation of signs of acute stress may be relatively straightforward, signs of chronic stress such as stereotypic behaviour require further research before they can be incorporated into QoL measures. So far, there has been little attempt to integrate proxy assessment with objective measures. We recommend that this integration would be beneficial. Fundamentally, both approaches aim to describe and quantify aspects of some inner state of well-being, and it should eventually be possible to map each on to the other.