The literature on the two main models of addiction (dopamine-based positive reinforcement and stress-based negative reinforcement models) have made many important contributions to understanding this brain disorder. However, rarely has there been a comprehensive critique of the limitations of both models. This article seeks to resolve theoretical issues inherent to each model, as well as propose a more comprehensive psycho-neuro-endocrinological theory of addiction which reconciles important elements of both. We suggest that there is not only direct interaction of dopaminergic and stress systems throughout the addiction cycle, from initial use, via the abusing stage, to the endpoint of addiction, but that this interaction is present prior to initial use. A combination of genetic factors and/or experiences of adversity may result in a stress-triggered sensitisation of dopaminergic networks which is present before the onset of substance use, which cannot be explained solely in terms of dopaminergic (positive) reinforcement. Rather these processes are best explained by an allostatic model which reconciles aspects of both models of addiction and shows how dopamine/stress interactions become increasingly pathological in the addiction cycle. Our model suggests that chronic stress eventually creates baseline hypodopaminergic activity, but also prompts dopaminergic hyperactivity in cue reactivity. This is the neural marker of allostatic mechanisms observed at endpoint addiction. We propose a multi-circuit explanation of how this cumulative effect of stress increasingly impacts on dopaminergic networks of reward, affect, attention, memory and behavioural control. This revised model provides a useful frame of reference for further research and ultimately clinical practice.