Among all neuromodulatory systems, the noradrenergic system is probably the one most closely related to vigilance and autonomic arousal (Aston-Jones et al. Reference Aston-Jones, Shipley, Chouvet, Ennis, van Bockstaele, Pieribone, Shiekhattar, Akaoka, Drolet and Astier1991; Berridge & Waterhouse Reference Berridge and Waterhouse2003; Carter et al. Reference Carter, Yizhar, Chikahisa, Nguyen, Adamantidis, Nishino, Deisseroth and de Lecea2010; Foote et al. Reference Foote, Aston-Jones and Bloom1980; Jacobs Reference Jacobs1986; Sara & Bouret Reference Sara and Bouret2012). The activity of locus coeruleus (LC) neurons is so closely related to arousal that autonomic measures such as pupil diameter are used as a proxy for LC activity in human studies (Einhäuser et al. Reference Einhäuser, Stout, Koch and Carter2008; Jepma & Nieuwenhuis Reference Jepma and Nieuwenhuis2011; Nassar et al. Reference Nassar, Rumsey, Wilson, Parikh, Heasly and Gold2012; Preuschoff et al. Reference Preuschoff, Marius't Hart and Einhäuser2011; Sterpenich et al. Reference Sterpenich, D'Argembeau, Desseilles, Balteau, Albouy, Vandewalle, Degueldre, Luxen, Collette and Maquet2006; Varazzani et al. Reference Varazzani, San-Galli, Gilardeau and Bouret2015). Much less clear, however, are the nature of the influence of LC activation–noradrenaline (NA) release on its targets in the brain and its implication for cognition. The GANE (glutamate amplifies noradrenergic effects) theory developed by Mather et al. is addressing this issue directly. This theory covers numerous aspects of cognition, ranging from attention and decision making to memory and emotions, and proposes an original cellular mechanism.

In line with earlier theories of NA functions, GANE emphasizes the effect of NA on gain, which presumably mediates the inverted-U-shaped relation between the efficacy of sensorimotor functions and arousal (Arnsten Reference Arnsten2009; Aston-Jones & Cohen Reference Aston-Jones and Cohen2005). Theories such as network reset and unexpected uncertainty are based on a very distinct intuition: The key role of the LC–NA system is to change internal representations, rather than enhance them (Bouret & Sara Reference Bouret and Sara2005; Yu & Dayan Reference Yu and Dayan2005). network reset is based on two features of the NA system: It is extremely well conserved across all vertebrates, and its activation is systematically associated with a profound change in behavior (Bouret & Sara Reference Bouret and Sara2004; Clayton et al. Reference Clayton, Rajkowski, Cohen and Aston-Jones2004; Dalley et al. Reference Dalley, McGaughy, O'Connell, Cardinal, Levita and Robbins2001; Devauges & Sara Reference Devauges and Sara1990; Jacobs Reference Jacobs1986; McGaughy et al. Reference McGaughy, Ross and Eichenbaum2008). The typical condition of LC activation is the orienting response to a salient stimulus (Aston-Jones & Bloom Reference Aston-Jones and Bloom1981; Bouret & Sara Reference Bouret and Sara2004; Foote et al. Reference Foote, Aston-Jones and Bloom1980). Unexpected uncertainty is based on similar intuitions and emphasizes the role of NA in learning (Yu & Dayan Reference Yu and Dayan2005). Again, there are some differences between neurobiological intuitions proposed in GANE versus network reset, but the key question the authors raise is not “how,” but “why”: “Why phasic arousal induced when encountering emotional stimuli can enhance processing of preceding stimuli when they have high priority” (sect. 8.2).

That question implies two features: First, the processing of stimuli is taking enough time to allow subsequent emotional stimuli to induce enhancement of this processing via an increase in arousal. This assumption makes strong predictions on the dynamics of these processes, and indeed, such an assumption is important in understanding LC/NA functions. Second, the processing of the original stimulus is not altered qualitatively by the emotional stimulus; it is only enhanced. In other words, the “priority maps” are not modified qualitatively after the onset of this emotional stimulus; arousal is only enhancing their impact on behavior.

The assumptions underlying network reset are different: A salient stimulus would cause a qualitative change in stimulus processing, both the nature of the representation and the associated neuronal activity. The priority maps would be changed. In the extreme version of network rReset, the highest priority would be attributed to the salient stimulus, and the preceding stimulus would be ignored. But if the initial stimulus leaves a trace strong enough to be integrated with the emotional one, after the reset, the new “functional network” would underlie the processing of both stimuli. In that case, the representation of the original stimulus would be modified (changed qualitatively), not quantitatively (enhanced or decreased).

Is the influence of arousing events qualitative (network reset) or quantitative (GANE)? Using the example in Figure 7 of Mather et al., these two theories make radically different predictions: According to network reset, the booming sound of a thunderstorm would not enhance the processing of the cow; it would first trigger an orienting response that consists of interrupting existing activity (including processing of the cow) and promoting redirection of attentional resources. Using the words of Mather et al., the sound would become “high priority,” but it would either be processed alone or be combined with the cow in a novel representation. Importantly, the representation of the cow as it existed before the storm would disappear.

Thus, we could rephrase Mather et al.'s question: Why should salient stimulus enhance processing of past events? First, time goes one way only, and modulating past events makes sense only if they are used for the present or for planning future actions (James Reference James1913; Sara Reference Sara2000). The example provided in Figure 7 is very close to laboratory situations in which discrete stimuli are manipulated in a controlled setting. But imagine yourself walking in the fields, and let's assume that for some reason, you are considering the cow. What will happen if you hear booming thunder? Will you still care about the cow? If yes, what is the chance that you think about it the same way you did before, independently of the critical information provided by the thunder? If the NA system had evolved to enhance the processing of the cow when an inherently more significant stimulus occurs, would this system be so widely represented among living animal species?

In conclusion, in addition to its influence on sensorimotor functions, the LC/NA system has a major role in promoting changes in behavior. The details of the model, including its dynamics, will be critical to understanding how, and why, the release of NA modulates forebrain systems. But this model should account for critical biological features of the LC: It is activated when a behaviorally relevant stimulus triggers a sympathetic response and a behavioral response. For all vertebrates, this autonomic activation is a generic emergency reaction that facilitates coping with a challenge (threat, effort, unexpected event, etc.), and it presumably facilitates the behavioral adjustment to the challenge. This adjustment may take several forms, including gain and/or reset, and be mediated by myriad neurobiological processes, but to understand why the central NA system exists and what it does, it is important to consider ecological problems that the brain has evolved to solve.