One major theme in studies of the neural basis of addiction has been a search for a final pathway common to the actions of all addictive drugs. For example, it has been proposed that addiction is caused by aberrant associative learning arising from altered dopamine neurotransmission (Di Chiara Reference Di Chiara1999; Kelley Reference Kelley2004a; Robbins & Everitt Reference Robbins and Everitt1999). Psychostimulants, as well as opiates, nicotine, ethanol, and cannabinoids, all have effects on dopaminergic neurotransmission through various pathways (Di Chiara & Imperato Reference Di Chiara and Imperato1988; Tanda et al. Reference Tanda, Pontieri and Chiara1997; Wise & Bozarth Reference Wise and Bozarth1987). Observations such as these have bolstered the hypothesis that addiction is at its root a unitary phenomenon. Indeed, even pathological gambling, an addiction that is not a substance dependence, may be related to changes in dopamine function (Zack & Poulos Reference Zack and Poulos2004). Of course the dopamine-learning hypothesis of addiction is only one account; other unified hypotheses have also been advanced to explain addiction. These include proposals that addiction results from disrupted hedonic homeostasis (Koob & Le Moal Reference Koob and Le Moal1997), from sensitization of incentive motivation (Robinson & Berridge Reference Robinson and Berridge2000), or from disruptions to prefrontal-mediated executive processes (Schoenbaum et al. Reference Schoenbaum, Roesch and Stalnaker2006a; Volkow & Fowler Reference Volkow and Fowler2000), among others. What these disparate theories have in common is that they each seem to suggest that addictive behavior arises from one basic dysfunction. But while each of these theories can marshal persuasive evidence in its support, each also has substantial areas of evidence that it cannot explain. For example, dopamine transporter knockout mice will still self-administer cocaine (Rocha Reference Rocha2003), an observation difficult to explain if altered dopamine neurotransmission is the only pathway to cocaine addiction.

In the current review article, Redish et al. have reconceptualized addiction as stemming not from a single unitary dysfunction but rather from a variety of underlying causes, each of which impact decision-making. In so doing, the authors take a position similar to that of others, who have viewed addiction as a disorder characterized by maladaptive decision-making (Bechara Reference Bechara2005; Kalivas & Volkow Reference Kalivas and Volkow2005; Schoenbaum et al. Reference Schoenbaum, Roesch and Stalnaker2006a). In this view, addictive behavior is that which has escaped the normal mechanisms that control decision-making. However the current review is unique in that it uses this idea to explore the many different aspects of decision-making that theoretically could be disrupted in addiction. Because decision-making arises from a collection of interacting processes, each with a distinct neurobiological basis, addiction could arise from a disruption to any one of these processes. After laying out the possible “vulnerabilities” in the decision-making process, the authors go on to show how dysfunction in each of these different vulnerabilities might be best explained by a different theory of addiction that has been advanced by other investigators. As this approach recognizes, the wide variety of theories of addiction that have been advanced over the past decades may not be mutually exclusive. Rather, each one could point to different members of a whole family of addictions.

The great strength of this approach is that it can be used to explain why addictions to different drugs, and even to a single drug in different cases, may have different properties. For example, multiple subtypes of alcoholism, which vary according to the amount and importance of craving, have been recognized by some clinicians (Addolorato et al. Reference Addolorato, Leggio, Abenavoli and Gasbarrini2005; Monterosso et al. Reference Monterosso, Flannery, Pettinati, Oslin, Rukstalis, O'Brien and Volpicelli2001). According to the authors' schema, addictions in which craving plays a large role could be explained by a different kind of dysfunction than those in which craving plays less of a role. However, even if certain theoretical distinctions, such as this one, turn out not to be applicable to addiction, the overall idea of distinguishing addictions by the specific aspect of decision-making that they disrupt is surely a good one. Mechanisms that appear to point to a final common pathway of all addiction, such as an increase in dopamine transmission in the ventral striatum, could be epiphenomena in some addictions rather than their root cause.

The ideas of Redish and colleagues also raise some broad questions. One of these, for instance, is how the idea that addiction reflects multiple underlying decision-making dysfunctions could itself be tested. The very strength of the framework advanced by these authors – that it is so powerful and wide-ranging, and can subsume so many theories of addiction – is also a potential weakness. What kind of result could falsify it? Related to this is the question of whether all of the theoretical vulnerabilities in the decision-making process identified by Redish et al. are equal in explanatory power, or whether some of them are more fundamentally involved in addiction than others. One possibility might be that some apparently disparate vulnerabilities in decision-making could reduce to a common neurobiological substrate. For instance, the authors identify the overvaluation of a predicted outcome as one vulnerability that may be associated with dysfunction of the orbitofrontal cortex (OFC) in certain addictions. However, a drug-induced dysfunction of the OFC could also lead to other vulnerabilities separately identified by the authors, such as an abnormally fast discounting function (Roesch et al. Reference Roesch, Takahashi, Gugsa, Bissonette and Schoenbaum2007). The underlying problem in both of these vulnerabilities might be a disruption in the ability to signal the value of expected outcomes by the OFC. Such a finding would in no way invalidate the approach of Redish et al., but it might modify their framework. Presumably, these are the kinds of questions and hypotheses that this important new approach is designed to stimulate and bring to the forefront.