The authors propose a new dualist hypothesis regarding decision making in humans as a function of the balance between reward/adversity previous knowledge and expectations. They defend the existence of an adversity processing circuit (APC) including the dorsal anterior cingulate cortex (dACC), adjacent parts of the lateral orbitofrontal cortex, and projections to lateral habenula and ventral striatum. According to the author’s view, these interconnected areas are involved in both learning of harmful situations, risk assessment and active avoidance strategies. This circuit could be the alter ego of the reward-oriented circuit that would include ventral ACC, medial orbitofrontal cortex and ventral tegmental area (VTA). A central role in this model is attributed to lateral habenula that is connected to VTA and may have a direct inhibitory effect on its activation. The whole construction also includes dopamine and serotonin with their well-known role in reward reinforcement (for the first) and anxiety relief (for the second). What is proposed in this manuscript is a go/no-go approach of the human behaviour with dopamine increase (D1 loop) acting as a potent inhibitor of the APC in dACC whereas dopamine decrease biases the choice selection towards inhibitory avoidance. In a parallel line, serotonin increase is thought to inhibit dACC attenuating adversity processing. Last but not least, there would be a reciprocal inhibition, the APC activation leading to decreased dopamine and serotonin levels via the activation of lateral habenula that in turn favour further APC activation. The limits of this reciprocal inhibition are not discussed since at this stage no homeostatic mechanism is proposed.
This theoretical model represents a good starting point for further studies aiming to explore its validity. Some arguments surrounding the model are supported by fMRI studies and animal data, other clearly not. What renders even more complex this puzzle is the need to separate learning effects (referring to conditioning), from decision making and error assessment. To be the simplest possible, human action facing unpleasant stimuli or experiences includes a probabilistic dimension (assessing the risk), a strategic dimension (decision making based on probabilities), a learning component (accumulated experience that favours avoidance) and most importantly, a posteriori critical reassessment and disengagement from erroneous beliefs or strategies. Limiting the mostly unknown interplay between cortical and subcortical regions to a dual APC/RPC equilibrium is certainly attractive but fatally over-simplistic.
A solid body of recent evidences supports the idea that reward reinforcement and learning takes place in lateral OFC whereas medial OFC is mostly involved in decision making (see the work by Noonan et al., European J Neurosci 2012;35, PNAS 2010;107). In the same line, lateral OFC is implicated in reversal learning (changing his own mind, Hampshire et al Neuroimage 2012;59). In this context, ACC (both dorsal and ventral parts) are thought to mediate the translation of decision making into motor behaviors but also choice prediction and reward guided learning (Cai et al., J Neurosci 2012;32, Noonan et al., J Neurosci 2011;31). From this point of view, the focus of the article on ACC as a key player in prediction of reward and adaptive behaviour is correct. Another peace of evidence supporting (at least partly) the author’s view comes from recent fMRI studies on error processing. Dorsolateral ACC is involved in error active avoidance or correction (Hochman et al., PLOS One 2014;9) but also in error processing (including the understanding of errors). However, ACC is not the only player and the distinction between ventral and dorsal parts proposed by the authors remains quite speculative (both dorsal and ventral part are activated in reward-related paradigms but also in social exclusion as nicely demonstrated by Moor et al., Neuroimage 2012;59). In fact, a very recent study identified at least 40 error-related areas indicating that a large scale of circuits may be alternatively activated as a function of the emotional significance and repercussions of an error (Neta et al., J Neurosci 2015;35). This rich, informative but also contradictory literature should be discussed when dealing with the role of dACC in the management of negative experiences. A simple example; the error-related negativity is not strictly related to the dACC as suggested by the authors. As recently showed it is generated in posterior cingulate cortex (Agam et al., PLOS One 2014;9). The reference of anterior insula as an important region in pain, disgust and aversion experiences is also supported by the literature as well as its strong connectivity with ACC. In contrast, ventral striatum, habenula and mediolateral thalamus are also part of the reward guided learning circuits and should not be exclusively seen as subcortical mediators of negative experiences (Noon et al., J Neurosci 2011;31).
What dampens my enthusiasm for this well constructed manuscript is the attempt to add a neurochemical component via the dopamine/serotonin systems. The fact that serotonin attenuates anxiety, depression, worries and the presence of high densities of 5HT1a receptors in dACC (as in several other limbic areas) do not imply a pivotal role in decreasing the dACC activation (as an hypothetical starting point of the decreased firing of the ARC). Similarly the fact that dopamine release is a pivotal event in reward and addictive behaviours do not signify that this action depends on the dACC. To the best of my knowledge, no experimental evidence in animal models (primates) supports this idea. Clinical experience does not support this viewpoint either. Depressed patients with presumably low serotonin levels have frequently addictive behaviours with presumably low ARC finding and total absence of inhibitory avoidance. I’m afraid that real life is by far more complex.
The text includes at least 12 repetitions of the main theory (frequently with similar sentences). I would prefer to read how this experimental paradigm (that is still noteworthy) could be tested. The authors refer to optogenetics (a good option limited to animal models). They should propose activation paradigms (fMRI) or lesional models to test their hypothesis.