The various perspectives on addiction, as listed in Table 1 of the target article, remind me of the fabled blind men touching the elephant, and Redish et al. deserve credit for combining the different accounts to give us a sketch of the elephant's shape. But keen sight they have not given us, and in our attempts to understand addiction we are still groping in the dark.
Much of the article is a straightforward summary of various decision systems in the brain, each associated with distinct learning capacities; little of it covers actual work on addiction. But this is no fault of the authors, since little research so far has focused on the effects of addiction on neural systems mediating decision-making. Although recent studies have begun to look at this question (Nelson & Killcross Reference Nelson and Killcross2006; Schoenbaum & Setlow Reference Schoenbaum and Setlow2005; Stalnaker et al. Reference Stalnaker, Roesch, Franz, Burke and Schoenbaum2006), at present we must rest content with the hypothetical nature of the vulnerabilities enumerated.
In their review of the literature on the brain's decision systems, Redish et al. underscore the importance of behavioral analysis in our attempt to understand addiction, because every system they discuss is delineated by behavioral assays developed over many decades. Unfortunately, though such assays have become standard tools among serious students of learning and behavior, they are rarely used in addiction research.
In this age of molecular supremacy, what some researchers want, above all, is the so-called addiction center of the brain, so they can safely disregard the rest in their attempt to nail down the critical molecule(s) for addiction, in the hope that the molecule of their choice will turn out to be the master switch. Such wishful thinking is certainly understandable, but the danger with wishful thinking, in science as in life, is that sooner or later it will be shattered by reality. For decades, the nucleus accumbens served as the center of addiction, but as we learn more and more about the Pavlovian and instrumental learning systems of the brain, it is becoming clear that the accumbens (or the mesolimbic pathway) cannot be the sole substrate of addiction. If Redish et al. tell us anything about the neural mechanisms of addiction, it is that the entire brain (literally the cerebrum) is critical, but that distinct functional systems within it are involved in distinct “vulnerabilities.”
To some this is bound to be disappointing news, leaving little room for real science. To those who believe that science gains rigor as it zooms in on the particular, the integrative activity of the brain is no doubt too nebulous an object for any rigorous analysis. For others it may suggest a new approach that focuses on interactions between neural systems in behaving organisms.
How can the entire brain be tackled? Paradoxically, this is not a question to which neuroscientists today devote much thought, but we can hardly afford to ignore it any further, especially when attempting to understand a process like addiction. It is to precise behavioral analysis that we must turn if we wish to dissociate global functional processes at the level of neural circuits. And what is striking about the field of addiction is precisely the lack of any detailed characterization of behavior – of what animals actually do after they become addicted. There is no lack of data, but a lack of meaningful data on the effects of addictive substances on the decision systems in the brain, because the field remains fixated on a few uninformative measures like conditioned place preference. Such procedures are a legacy from the study of learning decades ago, and they do not reflect the critical developments in this field since then. Our knowledge of the decision systems reviewed by Redish et al. are largely the product of modern behavioral work, and they have yet to be examined in any detail in the field of addiction.
That is not to say that what we have learned about the various decision systems will provide an infallible guide for addiction research. Far from it; as this article makes clear, much conceptual confusion still remains regarding how these systems are to be demarcated. To me the categorization of various types of learning without a detailed discussion of specific data is not a fruitful approach. One can always find studies in the literature that happen to support a particular classificatory scheme. One lab showed X, another Y – but the key question is whether they were trying to measure the same thing and whether the experimental results can be replicated by others. In the absence of a general consensus about what is actually done and what the data are, given a defined condition, one can argue ad infinitum about which system is which, and how each should be named, without contributing anything to the debate.
In short, although Redish et al. provide a fine introduction to the study of addiction, and a convenient starting point for future discussions, their article also reveals how far we are from anything approaching genuine understanding. As so often happens in the primitive stage of scientific inquiry, a wild abundance of data masks the lack of fundamental experimental findings, and a great variety of perspectives hides the poverty of theoretical advances. This is due, no doubt, to our ignorance of the detailed mechanisms underlying reward and decision-making, and an article like this will have served its purpose if it can only remind us of the need to go back to the laboratory.
