Relationships between dogs and humans are unique in their diversity across time and place, a testament to the incredible flexibility of the canine genome. Rarely, however, has any animal species been elevated in status to the extent observed for dogs in some contemporary societies. Modern human-dog relationships are unique in evolutionary terms and remain poorly understood. In this paper it will be proposed that modern dogs fulfil roles different from, but equally important to, those played by their ancestors when they gathered our sheep and protected our prized possessions. These roles might best be conceptualised as satisfying the otherwise unmet psychological needs of human owners for, among other things, companionship and social support, a parental attachment figure, a dependent child figure, a therapist, spouse and witness to our existence. Once the work of dogs is defined in this way, questions about their ability to adequately perform these roles can legitimately be asked, as can questions about whether our current demands on dogs have influenced canine behaviour and welfare. This is clearly the case and the outcomes are not always positive. Canines with overt behavioural disorders, perhaps indicative of underlying psychopathology, now occupy a range of professionals with job descriptions that did not exist several decades ago. Welfare shelters overflow with dogs discarded by owners whose expectations, often unrealistic to begin with, are not met. Genetic diseases and morphological traits that adversely affect canine health, welfare and behaviour are increasing in prevalence. To ensure that both humans and dogs benefit to the fullest extent possible from our shared activities, applied scientists need to work more closely with theoretical scientists to develop a multidisciplinary approach to describing, explaining and promoting optimal human-dog relationships.