Loss aversion is the idea that people experience losses more saliently than gains of equal valence.
(Tversky & Kahneman, 1991). Such behavior has been well explained by models such as prospect theory,
but often causes humans to act in a game-theoretically irrational way. Studies show that a potential gain
often needs to be twice as large as a potential loss in order for the average person to accept a bet at even
odds (Thaler, Tversky, Kahneman, & Schwartz, 1997). This behavioral quirk is often thought to be
connected to our negativity bias, having saved us in the past from dangerous decisions (Rozin &
Royzman, 2001). However, nowadays, loss aversion often means we make worse decisions in the long
run than we otherwise would. It is this difference in our actual vs normative behavior that makes the topic
so interesting to researchers.
Loss aversion is typically studied by asking people a series of questions and mapping their
preferences. Common questions are variations of: “would you agree to a bet where you have a 50%
chance of winning 50 dollars, and a 50% chance of losing 30 dollars?” The sizes of the potential gain and
potential loss, along with their respective probabilities, all impact a person’s decision to accept or reject
the bet, though the responses do not tend to align with the expected value of the bet itself. Recently,
neuroscientists have been using fMRI to observe subjects as they respond to such bets, giving new insight
into which regions of the brain are activated and to what extent. Unfortunately, fMRI can only provide
correlational, not causal evidence, leaving much ambiguity (Rick, 2010).
Many studies have suggested that the amygdala plays a role in loss aversion, though there is no
consensus about what exactly the amygdala’s role is, or how large that role may be. The amygdala is located in the frontal temporal lobe, and has been implicated in many functions including understanding
and responding to emotions such as fear, and controlling the brain’s reward system (Aggleton, Everitt,
Cardinal, & Hall, 2000). It makes intuitive sense that the amygdala would play a role in this
systematically irrational behavior, as we often link irrationality with our emotions. Debates remain about
whether it is the sole region responsible for our peculiar decision patterns, or just one of several in a more
There are not currently many studies that directly investigate the role of the amygdala in loss
aversion, but we can piece together findings from related studies to get a better understanding of what
causes our behavior.
Evidence for the Amygdala as the Main Region Involved in Loss Aversion:
One set of studies paints a picture to jointly suggest that loss aversion is closely linked to fear
responses and, as such, the amygdala plays a large, essential role in loss aversion. These studies look at
how loss aversion is impacted by 1) emotion regulation, 2) amygdala lesions in monkeys, and 3)
amygdala lesions in humans. These studies together make it seem that amygdala is the main region
responsible for loss aversion.
A 2001 study observed monkeys with amygdala lesions and implicated the amygdala in inhibiting
non-social risk taking and reacting to threatening stimuli: functions closely related to loss aversion
(Prather et al., 2001). The amygdala helps monkeys interpret threatening cues and inhibit actions that
would be impacted by those cues. For instance, a healthy monkey would pause before reaching for an
apple right next to a snake, while a lesioned monkey would reach for it as quickly as if there were no
snake. The lack of the usual fear response could be seen as further implicating the amygdala in loss
aversion. Further, the study showed a dissociation in the lesioned monkeys between social fear and fear of
inanimate objects. Since loss aversion is generally considered in economic (i.e. non-social) settings, such a dissociation would not be surprising. However, this study did not focus on the types of decision that are
typically associated with loss aversion.
The idea that loss aversion has an emotional basis has continued to be reinforced (Camerer,
2005), so Sokol-Hessner, Camerer, and Phelps went a step further to examine how emotion regulation
techniques impact our behavioral and neuronal responses to financial losses (Sokol-Hessner, Camerer, &
Phelps, 2013). In the experiment, they told subjects to either view each bet on its own merits (the
“Attend” condition), or take a broader perspective of the bet as one of several they would be making (the
“Regulate” condition). Using fMRI to look at BOLD correlates, they discovered that an individual’s
success in regulating their choices is correlated with changes in amygdala responses to losses only, not to
gains. Further, they showed that behavioral loss aversion correlates with amygdala activity in response to
losses relative to gains. However, the study also found activation in the dlPFC, vmPFC, and striatum,
indicating that the amygdala doesn’t act alone in such decisions.
The most striking study providing evidence that the amygdala is what controls loss aversion
observed two patients with rare focal bilateral amygdala lesions. Each lesioned patient was compared to a
group of 6 healthy controls, matched for demographics. The lesion patients demonstrated virtually no loss
aversion (De Martino, Camerer, & Adolphs, 2010). Crucially, these patients also did not demonstrate an
increased desire for risk — they would rather have a 50% chance of winning $50 than a 30% chance of
winning $100. “The findings suggest that the amygdala plays a key role in generating loss aversion by
inhibiting actions with potentially deleterious outcomes.” However, a caveat with lesion studies in
humans is that the lesion may have produced the effect because it eliminated connectivity between other
relevant regions. It is also possible that the lesion included small pieces of other brain regions, beyond the
Evidence against the Amygdala as the Main Region Involved in Loss Aversion:
Several studies present findings that indicate maybe the roll of the amygdala as described above is
not completely accurate. A study showing that the amygdala may not process fear, as well as studies that
showed constant amygdala activity across various bets and evidence that reward-value coding does not
rely on the amygdala, counter the importance of the amygdala as we outlined before.
Studies of patient SM, who had bilateral amygdala damage, indicate that the amygdala does not in
fact power our fear response, but rather simply helps us know where to gather social cues (Adolphs et al.,
2005). In 1995, researchers showed SM a series of faces and found that she was unable to identify when a
person’s face expressed fear, despite having normal face-identifying ability (Adolphs, Tranel, Damasio, &
Damasio, 1995). In a follow-up study ten years later, researchers instructed SM to look at the eyes of the
faces. Much of our emotional expression is relayed through our eyes. Suddenly, SM was able to correctly
identify fearful faces (Adolphs et al,. 2005). The results indicate that we are able to identify fear even
without our amygdala, and that maybe its role is that of telling us where to look for social or emotional
Another study using fMRI seemed to provide evidence counter to the notion that the amygdala is
a major player in loss aversion (Tom, Fox, Trepel, & Poldrack, 2007). The researchers collected BOLD
data while subjects decided whether to accept or reject various bets, all with 50/50 chance of gaining or
losing, but with varying dollar amounts. The study found that “greater behavioral loss aversion was
associated with greater neural sensitivity” to both losses and gains (Tom et al., 2007). Several regions of
the brain were observed to have this association, but neither the amygdala nor the anterior insula (another
region implicated in emotion) were observed to have fire in a way that correlated with modulating loss
aversion. That being said, it’s possible that the lack of differing amygdala activity is due to the study
itself, and not the amygdala’s function. For instance, the range of potential losses may have been too
small to elicit a variety of responses in the amygdala ( De Martino, Camerer, & Adolphs, 2010 ).
Other regions that the amygdala projects onto, such as the orbital prefrontal cortex (OFC) and the
medial prefrontal cortex (MFC), still function even without the amygdala present (Rudebeck, Mitz,
Chacko, & Murray, 2013). These regions are known to help encode reward-value. Rudebeck and his team
trained monkeys to recognize different stimuli with different amount of fluid to be rewarded. The
monkeys were presented with two stimuli in a row then held a button of their choosing (almost always the
button associated with the high-reward stimulus) until the reward was delivered. The monkeys performed
this task before and after an amygdala lesion. Rudebeck’s team found a more significant decrease in OFC
activation after the lesion, compared to MFC. But, crucially, the OFC and MFC still were able to encode
reward value. This finding indicates that the amygdala may help amplify value encoding, but is not in fact
necessary for it to occur.
Having gone through several studies, it seems that the amygdala is necessary, but not sufficient
for loss aversion. De Martino’s 2010 study provides compelling evidence that loss aversion as a whole
cannot occur in the absence of the amygdala. However, the amygdala alone does not cause loss aversion.
There are two functions that combine to cause loss aversion: 1) value encoding: how much the
brain interprets a specific value to be, and 2) probability weighting: how likely the brain determines a
specific outcome to be. Rudebeck’s research made it clear that the amygdala is not necessary for value
encoding. However, it now seems that the amygdala’s main contribution to loss aversion is in the form of
probability weighting. This conclusion would also make irrelevant, for our purposes, debating whether the
amygdala is responsible for interpreting fears, or simply directing us to find social cues.
Probability weighting is hard to isolate, but one potential study could involve observing fMRI as
patients are presented with different probabilities of wins or losses. If the amygdala responds differently
to different framings of the same scenario (“30% chance of winning” vs “70% chance of losing”), and
also responds differently when presented the same probability but in separate cases (“30% chance of winning” vs “30% chance of losing), then there might be evidence that the amygdala does quick,
mathematically misleading calculations that lead to surprising probability weighting.
While we cannot know for certain where this debate will go next, it has already helped shape the
field of Neuroeconomics, and is sure to have a lasting impact.