Goal-directed learning and obsessive-compulsive disorder
Gillan & Robbins (2014)
Philosophical transactions of the royal society

Abstract

Compulsions in OCD may derive from manifestations of excessive habit formation. The irrational threat beliefs (obsessions) characteristic of OCD may be a consequence, rather than an instigator, of compulsive behaviour in these patients.

Introduction

Compulsivity is ‘a hypothetical trait in which actions are persistently repeated despite adverse consequences’.

There are two main schools of thought regarding the underlying mechanism that leads to compulsive behaviour. These are: 1) Cognitive, compulsivity is mediated by dysfunction in the assignment of value to available alternatives. The compulsive individual may view the cost of cessation of behaviour to be higher than the benefits of thereof. In this sense, the choice to continue the behaviour is purposeful and goal directed. Compulsions are performed to reduce the likelihood that an unwanted, or feared, consequence will take place. 2) OCD doesn’t necessarily arise form faulty value attribution or cognitive bias. It may result from goal-directed dysfunction that interacts with anxiety and irrational belief. Patients with OCD largely understand the relative value of the available outcomes and the cost of actions, and aim to promote expected values of outcomes and desist from compulsive behaviour, but cannot exert the necessary control over their actions to realise this goal.

Neurobiological parallels: habit and obsessive-compulsive disorder

Habits are responses that are automatically triggered by stimuli and are considered the functional reciprocal of goal-directed behaviours that are intentional and sensitive to the value of prospective goals.

Animals use both reflective and reflexive models of action selection. Goal-directed behaviour is more accurate, but requires effort and attention. It suffers in times of stress, perhaps as a result of increases in working memory load It subsides as we become comfortable with repetitive action following over-training of the stimulus-response pair, and when outcomes are less tightly coupled with responses.

A shift from associative to sensorimotor fronto-striatal circuits mediates the transition from goal-directed to habitual control over behaviour.

An important double dissociation between medial and lateral subregions of the dorsal striatum in the balance between goal-directed and habitual control over behaviour. Habitual responding can be induced in rodents by lesions of the dorsomedial striatum (caudatus). The drosomedial striatum is critical for goal-directed action control.

Disrupting activity in the dorsolateral striatum (putamen) preserves sensitivity to outcome value in rodents, even after extended training. The rodent prefrontal cortex there has been some suggestion of a similar double dissociation between the prelimbic and infralimbic cotices, with the former being associated with goal-directed learning, and the later necessary for the execution of habits.

The infralimbic, IL, may arbitrate between controllers, inhibiting the goal-directed system in favour of performing previously reinforced actions.

There is a key role for subregions of the ventromedial prefrontal cortex in goal-directed control over action.

Neurobiological changes associated with OCD have been identified within circuits that run form the formal lobes to the striatum, and via the direct and indirect pathways to the thalamus and back to the frontal cortex. The caudate nucleus and the orbital gyrus are the most consistent regions where OCD patients show abnormal patterns of functional activation.

Experimental evidence for goal-directed dysfunction in obsessive-compulsive disorder

OCD patients have a bias toward stimulus-response learning, at the expense of acquiring action-outcome associations.

Basic decision making processes are not affected in OCD patients.

OCD patients are proficient in their goal-directed control over responses prior to over-training. Their behaviour becomes excessively habit-based over the course of over-training. It is possible that avoidance habits are associated with impaired instrumental extinction in a general sense.

In OCD, there is a consistent shift in balance away from goal-directed associative control over action towards stimulus-response habits. Potential causes for this imbalance are: 1) OCD patients may have a deficit in action-outcome associative learning, which causes them to rely excessively on stimulus-response links that were previously reinforced. 2) Excessive stimulus-response learning in OCD might cause patients to lose their sensitivity to action-outcome links, producing deficits in explicit action-outcome knowledge Fundamental problem in action-outcome associative learning and/or execution exists in OCD, this is not dependent on excessive habit formation in the disorder. 3) The problem in OCD could lie in the arbitration between the controllers (model based and model free).

Habit biases in OCD are not necessarily driven by deficits in goal-directed contingency knowledge. It is plausible that both habit-based and goal-directed learning may be affected in OCD.

The model-based, model-free reinforcement-learning schema is a schema wherein model-based behaviour is hypothesized to map onto goal-directed action and model-free supports habit learning. It doesn’t consider the two system reciprocal, and can assess their independent contributions to choice behaviour.

Model-based control over action is selectively diminished in OCD patients.

Habits and compulsive-obsessive disorder (COD)

Based on recent observations, it is proposed that people have compulsions, and then have obsessions.

Two main problems with the current OCD framework ascribed by cognitive models of the disorder are: 1) Existing cognitive models rely on the supposition that obsessions drive OCD and compulsions are secondary phenomena But, in studies there is clear evidence that excessive compulsive-like automatic behaviours develop in OCD patients in the absence of any prior obsessions related to the task procedures. There is purely behavioural disturbance in OCD that is independent of obsessionality. 2) OCD is an ego-dystonic disorder, patients have insight in the irrationality of their actions.  

It is proposed that the behavioural disturbance is the critical component of OCD and it has its neurobiological biases in the circuits running between the OFC and the caudate.

In a model of COD atendency towards compulsive learning is central. The mechanism through which the other critical features of the disorder can be explained are both propagators and consequences of compulsivity.

Habits are associated with a hitherto unreported premonitory ‘urge to respond’. This urge may be critical in elaborating a model of how compulsive behaviour could give rise to obsessional thinking. Cognitive dissonance describes a state of conflict that arises when two or more competing beliefs are simultaneously held. Humans are motivated to reduce this conflict by altering one of these beliefs. This effect is observed in humans when behaviour contradicts belief. Irrational thoughts may be the product of the mind’s attempt to resolve the discrepancy between patient’s cognitions and their otherwise inexplicable urge to perform compulsive behaviours. The experience of the irresistible urge to perform, or the very performance of, compulsive avoidance behaviours may engender cognitive dissonance that is reconciled by the development of a new irrational belief about threat in the environment. This new fear makes sense of the need to compulsively perform avoidance responses and may contribute to the motivation of subsequent avoidance responding.

Because the compulsions in OCD are avoidant, when they are continually performed, the only demonstrable contingency the individual is exposed to is on where a state of safety follows the performance of an avoidance response. This prevents exposure to the crucial disconfirming case or extinction.

Stressors promote habit formation. Anxiety biases attention to stimuli, and away from outcomes, potentially causing a similar failure to execute goal-directed behaviours as seen during stress manipulations. While anxiety-related mechanism may contribute to habit biases in some patients, it is unlikely to fully account of them.

It may be the case that there are many routes to the OCD phenotype, and that dysfunction in habit learning and trait anxiety are independent, yet interacting diatheses.