Botulinum toxin-induced facial muscle paralysis affects amygdala responses to the perception of emotional expressions: preliminary findings from an A-B-A design

It has long been suggested that feedback signals from facial muscle activity associated with forming facial expressions are critical to emotional experience [1, 2]. This notion has developed into what we now know as the facial feedback hypothesis [3]. According to this hypothesis, our brain receives afferent feedback signals from facial muscles that significantly influence how we process and experience emotion [4]. Importantly, a number of psychological experiments have shown support for the facial feedback hypothesis [5‐9].

In a classic demonstration by Strack and colleagues, participants judged the funniness of cartoons while unknowingly contracting their zygomaticus (the facial muscle used during smiling) by holding a pen in their mouths [9]. This subtle manipulation influenced their perception of the cartoons; such that they found the cartoons to be funnier when contracting the zygomaticus muscles compared to a control condition where the zygomaticus muscles were not contracted. This study offered clear behavioral evidence that feedback signals from the facial muscles influence our emotional experiences.

Seminal studies such as this would be strengthened by additional work seeking to establish a direct causal relationship between facial feedback and emotional phenomena. Though Strack and colleagues convincingly employed methods that protected against demand characteristics (i.e., participants ostensibly were not aware of the fact that they were being asked to smile) [9], there are additional considerations when asking participants to actively alter their facial muscle contractions. Specifically, intentional alteration of facial muscle activity calls for novel efferent signals to be sent to the facial muscles, which might influence how we experience emotion independent of any change in the afferent facial feedback signals to the brain. To put it another way, this method does not allow us to distinguish between whether the observed outcomes are due to a change in brain activity that is responsible for producing the emotional facial expression (efferent) or a change in facial feedback signals to the brain (afferent) [10, 11]. Thus, it would be ideal to leave the efferent brain signals intact while selectively shutting down the afferent facial muscle signals.

Recent widespread cosmetic use of botulinum toxin (BTX) type-A has allowed researchers to perform exactly this separation of afferent and efferent signals, making it possible to directly test causal factors related to facial feedback. BTX induces a temporary paralysis of the muscles at the injection site by inhibiting the release of the neurotransmitter acetylcholine at the muscle nerves [12]. By injecting BTX into the muscles that we use to make emotional facial expressions, a “reversible lesion” of target facial muscles can be temporarily created. This effectively “severs” afferent feedback signals from these muscles while leaving the efferent signals intact, thus allowing us to study emotional experience and processing devoid of facial muscle feedback.

The corrugator supercilii—an important component of the glabellar muscles (the “frown muscles” between the eyebrows) along with the procerus [13]—has been implicated in the experience and processing of negative emotions. One prototypical feature of an angry facial expression is that the medial portions of the eyebrows are pulled down using the corrugator supercilii muscles [14]. Using facial electromyography (EMG), researchers have shown that corrugator activity is selectively potentiated when viewing photos of angry facial expressions [15‐17]. A similar pattern of corrugator activity is found in response to negative affective pictures [18‐20] and sounds [20].

Based on this established link between corrugator activity and negative emotion, a number of recent studies have used BTX to directly test the facial feedback hypothesis and found that feedback from facial muscles influenced subjective experience of emotion and emotional language processing [10, 21]. Other studies suggest that BTX injections mitigate depression symptoms and help elevate mood in major depressive disorder patients [22, 23]. Functional neuroimaging studies are also beginning to explore the relationship between facial feedback and responses to emotional stimuli. The first such study assessed the effect of BTX on the ability to imitate emotional expressions [24]. Amygdala activity, and its coupling with brain stem activity, was diminished when the corrugator muscle was paralyzed with BTX as participants attempted to imitate angry facial expressions. Taken together, these data suggest that our emotional experiences as well as brain activity to emotional stimuli can be influenced by feedback signals from the facial muscles.

The present study sought to determine whether the activity of the amygdala—a brain region within the medial temporal lobe that is known to be sensitive to facial expressions of emotion, including anger [25]—in response to angry facial expressions could be manipulated by BTX-induced glabellar (i.e., corrugator/procerus) muscle paralysis. Critically, when the effect of BTX injection subsides, would amygdala activity in response to angry facial expressions be restored? If so, this would provide strong evidence in favor of the facial feedback hypothesis and shed further light on its underlying neural mechanism. To this end, we used functional magnetic resonance imaging (fMRI) in an A-B-A design to assess neural responses to facial expressions a total of three times: prior to BTX treatment, shortly after BTX treatment (i.e., during corrugator/procerus paralysis), and after the effects of BTX had subsided. We hypothesized that amygdala activity would be greater to angry compared to happy facial expressions prior to BTX administration, that these signal increases to angry expressions would be mitigated by BTX, and that they would return to their pre-injection state (i.e., relatively enhanced amygdala responses to angry vs. happy faces) after BTX had subsided.

The aim of the current study was to investigate the effects of BTX-induced corrugator/procerus paralysis on amygdala activity to angry faces by adapting an A-B-A experimental design, a powerful method that allowed us to determine the neural consequence of BTX treatment. Consistent with our hypothesis, we observed diminished amygdala activity to angry vs. happy faces when the facial muscles used to produce angry facial expressions were paralyzed. Importantly, amygdala activity was restored to its pre-BTX injection state after the effects of BTX subsided. In other words, amygdala activity in response to angry vs. happy faces was modulated in parallel with the activity state of the corrugator supercilii and the procerus. When the corrugator and procerus muscles were unaffected by BTX, greater amygdala activity to angry vs. happy faces was observed. However, when the corrugator and procerus muscles were paralyzed via BTX injection, amygdala activity in response to angry vs. happy faces was attenuated. Our finding implies that feedback signals from the corrugator supercilii and the procerus to the brain are modulating amygdala activity.

According to the facial feedback hypothesis, our emotional experience is modulated by feedback signals from the facial muscles that we use to create expressions, and this argument is supported, at least in part, by a number of psychological studies [6, 8‐10, 21]. Temporary paralysis of facial muscles using BTX essentially cuts off all afferent feedback signals from the injection site while leaving the efferent signals untouched [12], providing researchers with an ideal setup to directly test the facial feedback hypothesis. Behavioral and psychophysiological studies utilizing these unique properties of BTX have also partially supported the facial feedback hypothesis [10, 21]. Here, we use the term “partially” because the reports from the existing literature make it quite clear that not all of our emotional experience can be accounted for by afferent facial feedback signals [6, 9, 10, 33]. At the same time, however, these studies do show evidence supporting the facial feedback hypothesis; that at least some of our emotional experience is influenced by afferent signals from facial muscles.

Our data add further support to the facial feedback hypothesis by offering direct neural evidence. We chose to focus on the amygdala, which is known to be responsive to biologically relevant and socially salient stimuli, including facial expressions [34]. We specifically selected angry faces to probe the amygdala, since the BTX treatment in our study targeted the corrugator supercilii and the procerus – the glabellar muscles that we use when we frown and make angry faces. Prior to BTX treatment, our participants demonstrated greater amygdala activity to angry faces compared to happy faces. Similar activity of the amygdala was observed when effects of BTX wore off at least 9 months after its injection to the corrugator supercilii and the procerus. It was only during the period when BTX was active and the corrugator and procerus muscles were paralyzed that we did not see greater amygdala activity to angry vs. happy faces. This quadratic pattern of amygdala activity is consistent with the aforementioned behavioral and psychophysiological evidence in favor of the facial feedback hypothesis. It is clear from the data that the lack of afferent feedback signals from the corrugator supercilii and the procerus has impacted how the amygdala responds to angry vs. happy faces. The causal relationship we observe here is as follows: if there are no afferent feedback signals from the facial muscles to the brain, then amygdala activity to angry vs. happy faces is diminished. We believe that this means amygdala activity to angry vs. happy faces is, at least partially, relying on the feedback signals from the corrugator supercilii and the procerus. This implies that when we see an angry face, we detect and experience negatively valenced emotion by contracting and flexing the relevant muscles—the corrugator supercilii being the most prominent one in this case—and our amygdala uses these afferent signals to properly process the information.

Our data showing diminished activity of the amygdala due to BTX-induced paralysis of the corrugator/procerus are consistent with the findings reported by Hennenlotter and colleagues [24]. In their fMRI investigation, they also observed attenuated amygdala activity to angry faces when the participants were treated with BTX and had their corrugator paralyzed, compared to their placebo-injected controls [24]. There were slight differences as well—for instance, they found BTX-induced attenuation of amygdala activity only when the participants were asked to imitate the facial expressions that were shown on the screen, as opposed to just viewing them passively. In our study, participants were not explicitly asked to imitate the facial expressions they were viewing on the screen, but we observed diminished amygdala activity nonetheless. This discrepancy could stem from many factors, among which include differences in the experimental paradigm that was used to probe the amygdala. Since we did not explicitly instruct the participants to either imitate or not imitate the facial expressions, they might have been less aware of their facial muscular movements and thus less resistant to spontaneous elicitations of their own facial expressions when viewing other people’s faces. This may have also influenced the difference in the observed laterality of the amygdala—that is, whereas Hennenlotter and colleagues report that BTX affected the left amygdala [24], we found the right amygdala to be modulated by BTX-induced paralysis of the corrugator supercilii. Despite these relatively minor differences, we believe our data are in the same vein as the findings by Hennenlotter and colleagues, in the sense that both provide neural evidence supporting the facial feedback hypothesis [24]. Furthermore, an important contribution of the current study is the addition of a third post-BTX condition. As the last ‘A’ in the A-B-A design, this third condition allows us to directly test a causal relationship between facial feedback signals and amygdala activity and suggests that amygdala activation changes due to BTX injections are not permanent but rather reversible.

Another aspect of the current study that warrants discussion is the inclusion of happy faces in the analysis. Happy faces were used as a direct comparison to angry faces to gauge the level of amygdala activity—a common strategy that has been employed in a number of fMRI studies investigating the effects of negatively valenced facial expressions, such as anger or fear [35‐37]. Upon closer examination of the data, our observations revealed that amygdala activity to happy faces actually showed an opposite quadratic pattern compared to angry faces. By subtracting amygdala activity to happy faces from angry faces and thus calculating amygdala activity to angry vs. happy faces, the U-shaped quadratic pattern became more prominent in our data. This inverted U-shaped pattern of activity to happy faces emerges as a topic of future inquiry, as BTX only had an effect on the corrugator supercilii and the procerus, not the zygomaticus major—the facial muscle we use to smile—but nevertheless had an impact on amygdala activity.

BTX-induced corrugator/procerus paralysis did not alter the participants’ perception of the valence of angry or happy faces. Regardless of BTX treatment, angry faces were consistently rated as being negatively valenced, and happy faces were consistently rated as being positively valenced. This was expected, as judging such clearly valenced facial expressions would not necessarily require feedback signals from the corrugator, given that the participants have prior experience and knowledge of the meaning of angry and happy faces. It is important to note that these behavioral data do not imply that afferent signals from the corrugator/procerus are not being used during processing of emotional facial expressions. Rather, these results hint at our ability to utilize multiple sources of information when making decisions on a given clearly valenced face, such that the lack of corrugator feedback signal is compensated by other factors including past experience and learned responses. Taking into account the EMG findings that show increased corrugator activity to angry facial expressions [15‐17], negative pictures [18‐20] and sounds [20], as well as other BTX-induced corrugator paralysis studies demonstrating behavioral effects such as impaired negative emotional language processing (i.e., significantly delayed reading times for sentences describing angry and sad situations) [21] and diminished subjective emotional experience in response to mildly positive video clips [10], it is clear that the feedback signals from the corrugator impact our emotional experiences, and it is beneficial to interpret our data within this context.

Limitations of the current study include the relatively small sample size. While within subject repeated measures (i.e., A-B-A design) mitigates this limitation to an extent, having a larger sample size, as well as a control group that is also cosmetically treated but without paralysis (see [10]) would strengthen the study. In addition, the present results might generalize only to females who would voluntarily report for BTX treatment. Next, it should be noted that the participants were exposed to surprised faces along with angry and happy faces during the experiment, and so the results reported here for happy and angry expressions could depend on the presence of surprised expressions in the experimental context. Finally, the current A-B-A design comprised uneven time intervals between each repeated measure, raising the possibility that the decrement in responses between pre-BTX and BTX (3–6 weeks) might be due to response habituation alone compared to the longer BTX and post-BTX interval (9 months). However, previous studies examining the test-retest reliability of amygdala BOLD activity to angry and fearful faces report reliable responsivity for both short (2 weeks; [38]) and long (1 year; [39]) time intervals.

To summarize, our data support the facial feedback hypothesis by offering neural evidence concerning the causal relationship between activity in the corrugator/procerus muscles and in the amygdala. Using BTX to temporarily paralyze facial muscles, we have found that the amygdala closely tracks the state of the corrugator/procerus by displaying a quadratic pattern of activity. Specifically, amygdala responses to angry faces were diminished by BTX-induced corrugator/procerus paralysis but reverted to pre-BTX injections levels once the effect of the drug wore off. These findings offer preliminary causal evidence that amygdala activity is sensitive to facial feedback during the perception of the facial expressions of others.