On habits and addiction: An associative analysis of compulsive drug seeking

Abstract

The processes that underlie the pathological pursuit of drugs in addiction and that support the transition from casual drug taking to their compulsive pursuit have recently been proposed to reflect the interaction of two action control processes that mediate the goal-directed and habitual control of actions for natural rewards. Here we describe the evidence for these learning processes, their associate structure and the motivational mechanisms through which their operation is translated into performance. Finally, we describe the potential changes in the interaction between habitual and goal-directed processes induced by drug addiction that subserve compulsive drug pursuit; i.e. the increase in habit learning and reduction in the regulation of habits induced by changes in the circuitry that mediates goal-directed action.

Figure 1

Illustration of contingency degradation and outcome devaluation procedures used to test whether actions are goal-directed in instrumental conditioning. Rats are trained on two actions, e.g. two levers, and rewarded with different outcomes for each action, here food pellets and sugar (left panel). After training two kinds of test are conducted: (i) The contingency degradation test (depicted in the top-center panel) in which one or other outcome is delivered non-contingently at the same probability as it is earned contingent on lever pressing. The other outcome continued to be earned only by lever pressing. (ii) An outcome devaluation test (depicted in the lower-center panel) prior to which one or other outcome is devalued either by sensory-specific satiety (top) or taste aversion learning (bottom). After each of these treatments rats are given a choice test in extinction (right panel) to assess the effects of the degradation and devaluation manipulations on choice.

Figure 2

Pavlovian-instrumental transfer. In this depiction, rats are first given pairings between three different auditory stimuli and three different food outcomes. Later they are trained on two actions, say two levers, delivering two of the three outcomes used in the first phase as depicted in the left panel of Figure 1. They are subsequently given a choice test on the two levers in which the three stimuli are presented in extinction. As previously reported (e.g. [29], [31]), this treatment generates evidence for two forms of transfer effect: (i) a general form, which is here depicted by the ‘maracas stimulus paired with a food outcome that is not then earned by lever pressing the presentation of which results in a general increase in the performance of both actions and (ii) a specific form, here depicted by the ‘bell’ and ‘tuning fork’ stimuli associated with the food outcomes that were also earned by pressing the levers. The effect of the stimulus presentation in this situation is to bias choice towards the action that, in training, earned the outcome predicted by the stimulus; e.g. if the left lever earned sugar then the tuning fork would enhance responding on that lever and not the bell, which would bias responding towards the right lever.

Figure 3

(a) Results of a devaluation extinction test (left panel) and punishment test (right panel) conducted after rats have been given undertrained (UT) or overtrained (OT) to lever press for sugar and then the sugar devalued by taste aversion learning. (b) Results from a choice outcome devaluation extinction test (left panels) and punishment test (right panels) in which rats have been trained to press two levers for distinct outcomes after being given either sham surgery or bilateral lesions of posterior dorsomedial striatum (pDMS). See text for details.