Emotional arousal enhances memory of currently relevant – that is, salient – information, whereas it can impair memory of irrelevant information (Bennion et al. Reference Bennion, Ford, Murray and Kensinger2013; Mather & Sunderland Reference Mather and Sutherland2011). Mather et al. formulate the interesting hypothesis that when norepinephrine (NE) release coincides with high glutamatergic activity within an activated brain region or neuronal ensemble, NE release is increased further, resulting in locally enhanced neuronal activity and better memory. In contrast, when NE release does not coincide with high glutamate levels, NE suppresses neuronal activity, resulting in memory impairment. Although their model incorporates interactions at the systems level, it places strong emphasis on local processes, creating NE “hotspots.” Here, we argue that such primarily local effects underestimate the importance of modulatory influences of the amygdala on encoding and consolidation of information throughout the network and that, without a functioning amygdala, such NE hotspots might be unable to affect local mnemonic processes.

According to the widely accepted “amygdala modulation hypothesis,” basolateral amygdala (BLA) activity enhances memory of emotionally arousing experiences by influencing neural plasticity mechanisms in target regions elsewhere (McGaugh Reference McGaugh2002). In rodents, pharmacologically enhancing or reducing noradrenergic activity within the BLA, that is, mimicking different arousal conditions, is sufficient to alter training-associated neural plasticity in distal brain regions (Beldjoud et al. Reference Beldjoud, Barsegyan and Roozendaal2015; McIntyre et al. Reference McIntyre, Miyashita, Setlow, Marjon, Steward, Guzowski and McGaugh2005) and to determine whether neural representations in these other areas are being strengthened (Roozendaal & McGaugh Reference Roozendaal and McGaugh2011). Recent evidence suggests that such BLA interactions with other brain regions not only modulate the strength of memory, but also are significantly involved in regulating memory precision (Ghosh & Chattarji Reference Ghosh and Chattarji2015), and that NE activity in particular may be the driving force behind improved accuracy (Barsegyan et al. Reference Barsegyan, McGaugh and Roozendaal2014). Human neuroimaging research corroborates these findings by showing that amygdala activity during encoding of emotionally arousing stimuli predicts enhancement of hippocampus-dependent memory (Canli et al. Reference Canli, Zhao, Brewer, Gabrieli and Cahill2000; Hamann et al. Reference Hamann, Ely, Grafton and Kilts1999). ß-Adrenoceptor blockade during encoding abolishes the emotional memory enhancement effect (Cahill et al. Reference Cahill, Prins, Weber and McGaugh1994) and suppresses memory-related amygdala activity (Strange & Dolan Reference Strange and Dolan2004). Amygdala–hippocampal connectivity, furthermore, is stronger for emotionally arousing than for neutral stimuli (Dolcos et al. Reference Dolcos, LaBar and Cabeza2004), and the dominant directionality of this connectivity is indeed from amygdala toward hippocampus (Fastenrath et al. Reference Fastenrath, Coynel, Spalek, Spalek, Milnik, Gschwind, Roozendaal, Papassotiropoulos and de Quervain2014).

Critically, amygdala–NE interactions selectively enhance memory for emotionally arousing as compared with neutral stimuli (e.g., Cahill et al. Reference Cahill, Prins, Weber and McGaugh1994). Mather et al. posit that the amygdala modulation hypothesis explains this selectivity in terms of a trade-off in which resources are shifted toward the emotional stimuli. However, recent findings indicate that there may be more to it than a simple trade-off. For example, Lovitz and Thompson (Reference Lovitz and Thompson2015) report that intra-BLA infusion of the β-adrenoceptor agonist clenbuterol induces a long-term increase in excitability of hippocampal neurons when administered after emotionally arousing inhibitory avoidance training, but that clenbuterol decreases hippocampal excitability in non-trained control animals. These findings strongly support the idea that the impairing effects of amygdala–NE interactions on memory of non-salient/non-arousing information involve an active process that is dependent on the amygdala.

Converging human evidence for this notion comes from patients with damage to the amygdala. For example, patients with Urbach–Wiethe disease (UWD), who exhibit selective calcifications in the BLA (Terburg et al. Reference Terburg, Morgan, Montoya, Hooge, Thornton, Hariri, Stein and van Honk2012), fail to show emotional enhancement of episodic memory (Cahill et al. Reference Cahill, Babinsky, Markowitsch and McGaugh1995). Furthermore, studies in patients with other forms of amygdala pathology revealed a deficit in upregulating processing of emotional stimuli in higher-order visual cortices (Vuilleumier et al. Reference Vuilleumier, Richardson, Armony, Driver and Dolan2004), as well as an impairment in increasing encoding-related hippocampal activity for emotional items (Richardson et al. Reference Richardson, Strange and Dolan2004). Critically, UWD patients also exhibit enhanced memory for neutral information encountered in close temporal proximity to emotionally arousing stimuli (i.e., diminishing the impairment for such information observed in healthy controls [Strange et al. Reference Strange, Hurlemann and Dolan2003]). One could argue that such findings remain consistent with an interpretation in terms of local hotspots of NE activity if amygdala damage would lead to a general impairment of NE signaling. However, UWD patients, although they fail to acquire conditioned responses, appear to exhibit normal arousal responses, as evidenced by normal skin conductance and startle responses to unconditioned stimuli (Bechara et al. Reference Bechara, Tranel, Damasio, Adolphs, Rockland and Damasio1995; Klumpers et al. Reference Klumpers, Morgan, Terburg, Stein and van Honk2015). Thus, findings from amygdala-lesioned patients agree with work in animals in suggesting that because of BLA damage, NE is ineffective in modulating local memory processes elsewhere in the brain.

Other studies have indicated that stress-related hormones such as glucocorticoids also contribute to selective enhancement of emotional memories. For example, in humans, elevating stress hormone levels after learning generally leads to consolidation benefits for emotionally arousing as compared with neutral information (Abercrombie et al. Reference Abercrombie, Speck and Monticelli2006; Kuhlmann & Wolf Reference Kuhlmann and Wolf2006). Work in rodents has shown that NE activity within the amygdala also crucially determines the modulatory effects of stress hormones on neural plasticity and memory in distal brain regions (Roozendaal et al. Reference Roozendaal, Nguyen, Power and McGaugh1999). The synthetic glucocorticoid dexamethasone, given immediately after inhibitory avoidance training, enhances long-term memory of this training in rats with an intact BLA, but dexamethasone impairs inhibitory avoidance memory if noradrenergic activity in the BLA is blocked with a β-adrenoceptor antagonist (Quirarte et al. Reference Quirarte, Roozendaal and McGaugh1997). Thus, these findings again support a critical role for BLA noradrenergic activity in determining enhancements or impairments of information storage in other brain regions.

In conclusion, local hotspots of NE activity at sites where mnemonic operations take place alone cannot explain the selectivity afforded by amygdala-driven modulatory processes. This observation, of course, raises the question of what mechanism underlies these distant modulatory effects. Important clues have come from functional connectivity studies in humans, showing that modulated regions are part of distinct large-scale neural systems, such as the “salience” and “default mode” networks (Hermans et al. Reference Hermans, van Marle, Ossewaarde, Henckens, Qin, van Kesteren, Schoots, Cousijn, Rijpkema, Oostenveld and Fernandez2011; Reference Hermans, Battaglia, Atsak, de Voogd, Fernández and Roozendaal2014). Novel technologies for electrophysiological recordings and optogenetics in rodents are beginning to make it possible to study such networks in unprecedented spatiotemporal detail. We predict that these developments will ultimately lead to the conclusion that selective processing of arousing material results primarily from amygdala-driven changes in network properties of large-scale neural systems, rather than NE-induced local hotspots of activity.