Essays on Science and SocietyESSAY

Parsing reward from aversion

See allHide authors and affiliations

Science  04 Nov 2016:
Vol. 354, Issue 6312, pp. 558
DOI: 10.1126/science.aak9762

Starting from the moment we hear our alarms in the morning, our emotions guide the thousands of decisions we make every day. More specifically, it is the valence of our emotions that determines our subsequent behavior. Valence is a concept that was originally defined in psychology and corresponds to the value we assign to the perceptions of our external and internal environments (1). Valence varies from negative, when we are afraid or anxious, to positive, when we are happy or peaceful. In the case of the morning alarm, if your emotional state has a positive valence you might jump out of bed, eager to engage with whatever is motivating you. Conversely, if your emotional state has a negative valence, you might choose to stay in bed to Avoid the causes of your negative emotions.

Valence circuitry

A proposed model for the role of basolateral amygdala projector populations in innate and learned valence


Beyond allowing or preventing your timely arrival to work, the neural circuits supporting valence encoding are critical for your survival. Imagine the consequences if you were to assign a positive valence to, and therefore seek, dangerous situations, for example. In fact, misassignment of valence is not rare, and dysfunctions of the circuits encoding valence are thought to underlie many psychiatric illnesses including anxiety, depression, addiction, and compulsive disorders (2, 3). From a fundamental and clinical point of view, understanding how the brain attributes valence to contexts and salient elements of our environment is one of the main challenges of modern neuroscience.

It has been known for decades that the basolateral amygdala (BLA) is necessary for associative learning of both positive and negative valence (4, 5). However, understanding how a single brain region can be responsible for encoding such a diverse range of information has been elusive.

One hypothesis to explain this paradox is that specific subpopulations of the BLA encode opposite valence. As a postdoctoral fellow in the Laboratory of Professor Kay Tye at the Massachusetts Institute of Technology (MIT), I have dedicated the past 4 years to testing this hypothesis, combining cutting-edge approaches in functional neuroanatomy, electrophysiology, and optogenetics. During this time, my collaborators and I assessed different subpopulations of the BLA that relay integrated information to distinct regions. Over multiple studies, I identified circuit, synaptic, and molecular mechanisms by which three distinct neural populations in the BLA encode innate and learned behaviors (68).

Consistent with recent studies in other brain regions (9), we found that different BLA neuronal populations (as defined by their projection targets) differentially encode valence. To do so, we used optogenetic-mediated photo-tagging in combination with large-scale electrophysiological recordings during the retrieval of positive or negative associative memories in mouse models and examined the activity of BLA neurons projecting to the nucleus accumbens, the centromedial amygdala, or the ventral hippocampus (see the figure).

Finalist: Anna Beyeler

Anna Beyeler received her undergraduate degree from the University of Bordeaux, in southern France, where she then completed her Ph.D. degree requirements. As a postdoctoral fellow at the Massachusetts Institute of Technology, she has been exploring the neural circuit mechanisms underlying rewarding and aversive memories. Dr. Beyeler is in the midst of establishing an independent research program aimed at identifying neural substrates of anxiety disorders at the University of Lausanne, Switzerland.


Although heterogeneity of response types was known to exist in the entire BLA, we were surprised to find that when mice encounter cues predicting a rewarding or aversive outcome (10, 11), heterogeneity of response types was also noted in BLA subpopulations. However, two BLA populations differentially encode valence during the retrieval of cues of opposite valence (7). Specifically, BLA neurons projecting to the nucleus accumbens were preferentially excited by cues predicting a reward, and those projecting to the central amygdala were preferentially excited by cues predicting an aversive outcome. The response-type heterogeneity we observed in BLA subpopulations could play a role in behavioral flexibility in cases in which the valence of environmental stimuli changes.

Consistent with our recordings in vivo, the synaptic inputs of BLA neurons projecting to the nucleus accumbens are potentiated after mice learn to associate a cue with a reward and depressed after mice learn to fear a cue associated with a foot shock (8). Conversely, the synaptic inputs of BLA neurons projecting to the centromedial amygdala were potentiated after fear learning and depressed after reward learning (8). Optogenetic activation of these projections induced approach and avoidance behavior, respectively (8), showing that these divergent pathways causally control valence encoding.

Although we had previously found that optogenetically simulating the BLA neurons that project to the ventral hippocampus drives negative valence in innate anxiety-related behaviors (6), we determined that these neurons do not preferentially respond to cues predicting unpleasant outcomes (7). These results add a level of complexity to our understanding of valence circuits and support a model of distinct routing for the attribution of valence for innate and learned information.

We have begun to unravel how BLA populations encode valence in physiological conditions (68). We can now start to investigate how valence circuits are dysfunctional in pathologies such as anxiety and depression. Ultimately, we hope this research will lead to the development of novel strategies to restore the function of those circuits in human patients.


Navigate This Article