The framework outlined by Redish et al. raises some important questions:

1. The ten vulnerabilities described can possibly be collapsed into fewer, more fundamental, vulnerabilities. Vulnerabilities 1 and 2 both infer that the perceived value of drugs increases in withdrawal (i.e., drug efficacy is enhanced by negative reinforcement). Vulnerabilities 3 and 4 are, in essence, positive reinforcement models. Vulnerabilities 4, 5, and 10 reflect a hyper-efficient learning process, such that drug-cue associations are learned quickly, are unusually strong, and the conditioned responses to such cues are resistant to extinction. Thus, the framework could be simplified by collapsing these vulnerabilities into fewer underlying processes. We make these observations based on the behavioural consequences of each vulnerability, which is not to say that each vulnerability does not have its own distinct neural substrate.

2. Vulnerability 9 relates to enhanced delay discounting. The authors refer to this as “impulsivity.” However, “impulsivity” has been defined by different authors in many different ways, some of which vary radically from delay discounting, for example, deficits in response inhibition (Vulnerability 8), lack of perseveration, lack of perseverance, boredom susceptibility, sensation seeking, functional impulsivity, reflection impulsivity, and so on. For example, a recent analysis of behavioural tests of impulsivity (Reynolds et al. Reference Reynolds, Ortengren, Richards and de Wit2006) indicated that impulsivity related to disinhibition (Vulnerability 8) can be dissociated from impulsive decision making related to delay discounting (Vulnerability 9). There is arguably a case for dispensing totally with the potentially misleading umbrella term “impulsivity.”

The current framework suggests that delay discounting is a stable characteristic (a trait). However, delay discounting rates are not stable in individuals, as they tend to increase during acute abstinence, as discussed further on (Field et al. Reference Field, Santarcangelo, Sumnall, Goudie and Cole2006; Giordano et al. Reference Giordano, Bickel, Loewenstein, Jacobs, Marsch and Badger2002).

3. The mechanism(s) involved in delay discounting (Vulnerability 9) are described as “a source of much controversy.” Delay discounting may critically involve the basic cognitive process of time perception (Wittmann & Paulus Reference Wittman and Paulus2007). Alterations in time perception might affect many vulnerabilities involving the planning system. For example, a slowing down of time perception in “impulsive” individuals might, perhaps counterintuitively, protect individuals from overvaluation of expected rewards (Vulnerability 3). This raises the important issue (discussed briefly by Redish et al) of interactions that occur among vulnerabilities. Our own work shows that nicotine withdrawal increases delay discounting (Field et al. Reference Field, Santarcangelo, Sumnall, Goudie and Cole2006), showing that Vulnerabilities 1 and 2 interact with Vulnerability 9 (delay discounting). Similarly, Verdejo-Garcia et al. (Reference Verdejo-Garcia, Bechara, Recknor and Perez-Garcia2007) have reported that, in individuals with substance dependence, the type of “impulsivity” best predictive of lifestyle (legal, employment, family, and social) problems is “urgency,” a tendency to respond impulsively when in negative emotional states (Whiteside & Lynam Reference Whiteside and Lynam2001). Such states would clearly be induced by drug withdrawal of various types. This again indicates that Vulnerabilities 1 and 2 interact with Vulnerability 9. Moreover, these findings suggest that such interactions may well be of considerable clinical relevance.

4. Real-world decisions about using drugs involve both delay and probability. The user makes a complex decision about the costs and benefits of using drugs in both the short- and long-term. The immediate gratification of using drugs is a certainty, whereas the benefits of abstinence are not. Focusing on delayed reward alone does not capture the essence of real-world decision-making.

5. The authors interpret attentional bias for drug-related cues in terms of Vulnerability 5. However, we suggest that, theoretically, attentional bias might also belong under Vulnerability 4 (overvaluation in the planning system/sensitization of motivation). For example, Franken (Reference Franken2003) suggested that attentional bias and craving both develop as a consequence of dopaminergic sensitization, and that the two have reciprocal causal effects on each other (craving increases attentional bias, and vice versa). Indeed, our research demonstrates that attentional bias is both associated with (e.g., Field et al. Reference Field, Mogg and Bradley2005), and caused by (e.g., Field et al. Reference Field, Mogg and Bradley2004) increases in subjective craving, and that direct manipulations of attentional bias can influence subjective craving (Field & Eastwood Reference Field and Eastwood2005). Attentional bias can also be understood in the context of Vulnerability 10 (a hyper-efficient learning process) because it develops rapidly as a consequence of pairings between the drug and drug-related cues (for review, see Hogarth & Duka Reference Hogarth and Duka2006).

6. Recent theoretical views (Goldstein & Volkow Reference Goldstein and Volkow2002; Wiers et al. Reference Wiers, Bartholow, van, Wildenberg, Thush, Engels, Sher, Grenard, Ames and Stacy2007) suggest potential interactions between the “salience” of conditioned drug-related cues (i.e., attentional bias for those cues; Vulnerabilities 4, 5, and 10), and deficient inhibitory control (Vulnerability 8). According to these models, inhibitory control mediates the impact of conditioned cues, either by directly suppressing responses to those cues, or by acting as a “brake” to prevent powerful responses to cues from influencing drug-seeking behaviour. These models are not inconsistent with the current framework, but they do suggest how Vulnerability 8 (selective inhibition of the planning system) can influence specific vulnerabilities within the planning system.

7. A radical implication of the theorising outlined in the target article is that individuals develop (relatively) unique paths to addiction, and that treatments based on an individual's specific vulnerabilities are required. Although this is an important idea, we wonder how this would work in practice (even if clinical tools could be designed to assess vulnerabilities), given that vulnerabilities are likely to interact, as outlined earlier. Interactions among vulnerabilities render accurate predictions from the framework difficult to make, particularly as they will change in any individual over time and with drug experience or drug withdrawal. We also wonder how receptive clinicians would be to a framework based on fundamental neuroscience.

In summary, we believe the framework outlined could possibly be both simplified and refined, particularly in terms of emphasising mechanistic interactions among vulnerabilities.