Natural rewards preferentially stimulate dopamine transmission in the nucleus accumbens shell. This effect undergoes adaptive changes (one-trial habituation, inhibition by appetitive stimuli) that are consistent with a role of nucleus accumbens shell dopamine in associative reward-related learning. Experimental studies with a variety of paradigms confirm this role. A role in associative stimulus-reward learning can provide an explanation for the extinction-like impairment of primary reinforcement that led Wise to propose the 'anhedonia hypothesis'. Addictive drugs share with natural rewards the property of stimulating dopamine transmission preferentially in the nucleus accumbens shell. This response, however, in contrast to that to natural rewards, is not subjected to one-trial habituation. Resistance to habituation allows drugs to activate dopamine transmission in the shell non-decrementally upon repeated self-administration. It is hypothesized that this process abnormally strengthens stimulus-drug associations thus resulting in the attribution of excessive motivational value to discrete stimuli or contexts predictive of drug availability. Addiction is therefore the expression of the excessive control over behaviour acquired by drug-related stimuli as a result of abnormal associative learning following repeated stimulation of dopamine transmission in the nucleus accumbens shell.