Rage Against the Machine: Advancing the study of aggression ethology via machine learning

The role of gene-environment interactions in social dysfunction: Focus on preclinical evidence from mouse studies

Human and animal research has demonstrated that genetic and environmental factors can strongly modulate behavioral function, including the expression of social behaviors and their dysfunctionalities. Several genes have been linked to pathologies characterized by alterations in social behaviors, e.g., aggressive/antisocial personality disorder (ASPD), or autism spectrum disorder (ASD). Environmental stimulation (e.g., physical exercise, environmental enrichment) or adversity (e.g., chronic stress, social isolation) may respectively improve or impair social interactions. While the independent contribution of genetic and environmental factors to social behaviors has been assessed in a variety of human and animal studies, the impact of their interactive effects on social functions has been less extensively investigated. Genetic mutations and environmental changes can indeed influence each other through complex mutual effects, e.g., inducing synergistic, antagonistic or interactive behavioral outcomes. This complexity is difficult to be disentangled in human populations, thus encouraging studies in animal models, especially in the mouse species which is the most suitable for genetic manipulations. Here we review the available preclinical evidence on the impact of gene-environment interactions on social behaviors and their dysfunction, focusing on studies in laboratory mice. We included findings combining naturally occurring mutations, selectively bred or transgenic mice with multiple environmental manipulations, including positive (environmental enrichment, physical exercise) and aversive (social isolation, maternal separation, and stress) experiences. The impact of these results is critically discussed in terms of their generalizability across mouse models and social tests, as well as their implications for human studies on social dysfunction.

Deep learning algorithms reveal increased social activity in rats at the onset of the dark phase of the light/dark cycle

The rapid decrease of light intensity is a potent stimulus of rats' activity. The nature of this activity, including the character of social behavior and the composition of concomitant ultrasonic vocalizations (USVs), is unknown. Using deep learning algorithms, this study aimed to examine the social life of rat pairs kept in semi-natural conditions and observed during the transitions between light and dark, as well as between dark and light periods. Over six days, animals were video- and audio-recorded during the transition sessions, each starting 10 minutes before and ending 10 minutes after light change. The videos were used to train and apply the DeepLabCut neural network examining animals' movement in space and time. DeepLabCut data were subjected to the Simple Behavioral Analysis (SimBA) toolkit to build models of 11 distinct social and non-social behaviors. DeepSqueak toolkit was used to examine USVs. Deep learning algorithms revealed lights-off-induced increases in fighting, mounting, crawling, and rearing behaviors, as well as 22-kHz alarm calls and 50-kHz flat and short, but not frequency-modulated calls. In contrast, the lights-on stimulus increased general activity, adjacent lying (huddling), anogenital sniffing, and rearing behaviors. The animals adapted to the housing conditions by showing decreased ultrasonic calls as well as grooming and rearing behaviors, but not fighting. The present study shows a lights-off-induced increase in aggressive behavior but fails to demonstrate an increase in a positive affect defined by hedonic USVs. We further confirm and extend the utility of deep learning algorithms in analyzing rat social behavior and ultrasonic vocalizations.

Simple Behavioral Analysis (SimBA) as a platform for explainable machine learning in behavioral neuroscience

The study of complex behaviors is often challenging when using manual annotation due to the absence of quantifiable behavioral definitions and the subjective nature of behavioral annotation. Integration of supervised machine learning approaches mitigates some of these issues through the inclusion of accessible and explainable model interpretation. To decrease barriers to access, and with an emphasis on accessible model explainability, we developed the open-source Simple Behavioral Analysis (SimBA) platform for behavioral neuroscientists. SimBA introduces several machine learning interpretability tools, including SHapley Additive exPlanation (SHAP) scores, that aid in creating explainable and transparent behavioral classifiers. Here we show how the addition of explainability metrics allows for quantifiable comparisons of aggressive social behavior across research groups and species, reconceptualizing behavior as a sharable reagent and providing an open-source framework. We provide an open-source, graphical user interface (GUI)-driven, well-documented package to facilitate the movement toward improved automation and sharing of behavioral classification tools across laboratories.

Independent inhibitory control mechanisms for aggressive motivation and action

Social behaviors often consist of a motivational phase followed by action. Here we show that neurons in the ventromedial hypothalamus ventrolateral area (VMHvl) of mice encode the temporal sequence of aggressive motivation to action. The VMHvl receives local inhibitory input (VMHvl shell) and long-range input from the medial preoptic area (MPO) with functional coupling to neurons with specific temporal profiles. Encoding models reveal that during aggression, VMHvl shellvgat+ activity peaks at the start of an attack, whereas activity from the MPO-VMHvlvgat+ input peaks at specific interaction endpoints. Activation of the MPO-VMHvlvgat+ input promotes and prolongs a low motivation state, whereas activation of VMHvl shellvgat+ results in action-related deficits, acutely terminating attack. Moreover, stimulation of MPO-VMHvlvgat+ input is positively valenced and anxiolytic. Together, these data demonstrate how distinct inhibitory inputs to the hypothalamus can independently gate the motivational and action phases of aggression through a single locus of control.

Effects of ketamine on rat social behavior as analyzed by DeepLabCut and SimBA deep learning algorithms

Traditional methods of rat social behavior assessment are extremely time-consuming and susceptible to the subjective biases. In contrast, novel digital techniques allow for rapid and objective measurements. This study sought to assess the feasibility of implementing a digital workflow to compare the effects of (R,S)-ketamine and a veterinary ketamine preparation Vetoquinol (both at 20 mg/kg) on the social behaviors of rat pairs. Historical and novel videos were used to train the DeepLabCut neural network. The numerical data generated by DeepLabCut from 14 video samples, representing various body parts in time and space were subjected to the Simple Behavioral Analysis (SimBA) toolkit, to build classifiers for 12 distinct social and non-social behaviors. To validate the workflow, previously annotated by the trained observer historical videos were analyzed with SimBA classifiers, and regression analysis of the total time of social interactions yielded R 2 = 0.75, slope 1.04; p < 0.001 (N = 101). Remarkable similarities between human and computer annotations allowed for using the digital workflow to analyze 24 novel videos of rats treated with vehicle and ketamine preparations. Digital workflow revealed similarities in the reduction of social behavior by both compounds, and no substantial differences between them. However, the digital workflow also demonstrated ketamine-induced increases in self-grooming, increased transitions from social contacts to self-grooming, and no effects on adjacent lying time. This study confirms and extends the utility of deep learning in analyzing rat social behavior and highlights its efficiency and objectivity. It provides a faster and objective alternative to human workflow.

Endocannabinoid system and aggression across animal species

This narrative review article summarizes the current state of knowledge regarding the relationship between the endocannabinoid system (ECS) and aggression across multiple vertebrate species. Experimental evidence indicates that acute administration of phytocannabinoids, synthetic cannabinoids, and the pharmacological enhancement of endocannabinoid signaling decreases aggressive behavior in several animal models. However, research on the chronic effects of cannabinoids on animal aggression has yielded inconsistent findings, indicating a need for further investigation. Cannabinoid receptors, particularly cannabinoid receptor type 1, appear to be an important part of the endogenous mechanism involved in the dampening of aggressive behavior. Overall, this review underscores the importance of the ECS in regulating aggressive behavior and provides a foundation for future research in this area.

Automatically annotated motion tracking identifies a distinct social behavioral profile following chronic social defeat stress

Severe stress exposure increases the risk of stress-related disorders such as major depressive disorder (MDD). An essential characteristic of MDD is the impairment of social functioning and lack of social motivation. Chronic social defeat stress is an established animal model for MDD research, which induces a cascade of physiological and behavioral changes. Current markerless pose estimation tools allow for more complex and naturalistic behavioral tests. Here, we introduce the open-source tool DeepOF to investigate the individual and social behavioral profile in mice by providing supervised and unsupervised pipelines using DeepLabCut-annotated pose estimation data. Applying this tool to chronic social defeat in male mice, the DeepOF supervised and unsupervised pipelines detect a distinct stress-induced social behavioral pattern, which was particularly observed at the beginning of a novel social encounter and fades with time due to habituation. In addition, while the classical social avoidance task does identify the stress-induced social behavioral differences, both DeepOF behavioral pipelines provide a clearer and more detailed profile. Moreover, DeepOF aims to facilitate reproducibility and unification of behavioral classification by providing an open-source tool, which can advance the study of rodent individual and social behavior, thereby enabling biological insights and, for example, subsequent drug development for psychiatric disorders.

Advancing social behavioral neuroscience by integrating ethology and comparative psychology methods through machine learning

Social behavior is naturally occurring in vertebrate species, which holds a strong evolutionary component and is crucial for the normal development and survival of individuals throughout life. Behavioral neuroscience has seen different influential methods for social behavioral phenotyping. The ethological research approach has extensively investigated social behavior in natural habitats, while the comparative psychology approach was developed utilizing standardized and univariate social behavioral tests. The development of advanced and precise tracking tools, together with post-tracking analysis packages, has recently enabled a novel behavioral phenotyping method, that includes the strengths of both approaches. The implementation of such methods will be beneficial for fundamental social behavioral research but will also enable an increased understanding of the influences of many different factors that can influence social behavior, such as stress exposure. Furthermore, future research will increase the number of data modalities, such as sensory, physiological, and neuronal activity data, and will thereby significantly enhance our understanding of the biological basis of social behavior and guide intervention strategies for behavioral abnormalities in psychiatric disorders.

Neurobiology of Aggression-Review of Recent Findings and Relationship with Alcohol and Trauma

Aggression can be conceptualized as any behavior, physical or verbal, that involves attacking another person or animal with the intent of causing harm, pain or injury. Because of its high prevalence worldwide, aggression has remained a central clinical and public safety issue. Aggression can be caused by several risk factors, including biological and psychological, such as genetics and mental health disorders, and socioeconomic such as education, employment, financial status, and neighborhood. Research over the past few decades has also proposed a link between alcohol consumption and aggressive behaviors. Alcohol consumption can escalate aggressive behavior in humans, often leading to domestic violence or serious crimes. Converging lines of evidence have also shown that trauma and posttraumatic stress disorder (PTSD) could have a tremendous impact on behavior associated with both alcohol use problems and violence. However, although the link between trauma, alcohol, and aggression is well documented, the underlying neurobiological mechanisms and their impact on behavior have not been properly discussed. This article provides an overview of recent advances in understanding the translational neurobiological basis of aggression and its intricate links to alcoholism and trauma, focusing on behavior. It does so by shedding light from several perspectives, including in vivo imaging, genes, receptors, and neurotransmitters and their influence on human and animal behavior.

Probing the relevance of the accelerated aging mouse line SAMP8 as a model for certain types of neuropsychiatric symptoms in dementia

INTRODUCTION

People with dementia (PwD) often present with neuropsychiatric symptoms (NPS). NPS are of substantial burden to the patients, and current treatment options are unsatisfactory. Investigators searching for novel medications need animal models that present disease-relevant phenotypes and can be used for drug screening. The Senescence Accelerated Mouse-Prone 8 (SAMP8) strain shows an accelerated aging phenotype associated with neurodegeneration and cognitive decline. Its behavioural phenotype in relation to NPS has not yet been thoroughly investigated. Physical and verbal aggression in reaction to the external environment (e.g., interaction with the caregiver) is one of the most prevalent and debilitating NPS occurring in PwD. Reactive aggression can be studied in male mice using the Resident-Intruder (R-I) test. SAMP8 mice are known to be more aggressive than the Senescence Accelerated Mouse-Resistant 1 (SAMR1) control strain at specific ages, but the development of the aggressive phenotype over time, is still unknown.

METHODS

In our study, we performed a longitudinal, within-subject, assessment of aggressive behaviour of male SAMP8 and SAMR1 mice at 4, 5, 6 and 7 months of age. Aggressive behaviour from video recordings of the R-I sessions was analysed using an in-house developed behaviour recognition software.

RESULTS

SAMP8 mice were more aggressive relative to SAMR1 mice starting at 5 months of age, and the phenotype was still present at 7 months of age. Treatment with risperidone (an antipsychotic frequently used to treat agitation in clinical practice) reduced aggression in both strains. In a three-chamber social interaction test, SAMP8 mice also interacted more fervently with male mice than SAMR1, possibly because of their aggression-seeking phenotype. They did not show any social withdrawal.

DISCUSSION

Our data support the notion that SAMP8 mice might be a useful preclinical tool to identify novel treatment options for CNS disorders associated with raised levels of reactive aggression such as dementia.

Neural circuits regulating prosocial behaviors

Positive, prosocial interactions are essential for survival, development, and well-being. These intricate and complex behaviors are mediated by an amalgamation of neural circuit mechanisms working in concert. Impairments in prosocial behaviors, which occur in a large number of neuropsychiatric disorders, result from disruption of the coordinated activity of these neural circuits. In this review, we focus our discussion on recent findings that utilize modern approaches in rodents to map, monitor, and manipulate neural circuits implicated in a variety of prosocial behaviors. We highlight how modulation by oxytocin, serotonin, and dopamine of excitatory and inhibitory synaptic transmission in specific brain regions is critical for regulation of adaptive prosocial interactions. We then describe how recent findings have helped elucidate pathophysiological mechanisms underlying the social deficits that accompany neuropsychiatric disorders. We conclude by discussing approaches for the development of more efficacious and targeted therapeutic interventions to ameliorate aberrant prosocial behaviors.

Excessive alcohol consumption after exposure to two types of chronic social stress: intermittent episodes vs. continuous exposure in C57BL/6J mice with a history of drinking

RATIONALE

The attraction to alcohol can be greatly increased when it is consumed in a social context. While pro-social interactions can potentiate voluntary alcohol drinking under some conditions, aversive social experience (i.e., social stress) can similarly intensify alcohol consumption.

OBJECTIVE

We sought to determine how exposure to different types of chronic social stress (i.e., intermittent episodes of social defeat or continuous social stress) influences alcohol consumption and the reinforcing effects of alcohol in mice with a history of drinking.

METHODS

Separate cohorts of male C57BL/6J mice were exposed to either 10 days of continuous or intermittent social defeat stress. In experiment 1, mice were assigned to 20% w/v alcohol consumption in a two-bottle choice protocol both prior to and after exposure to social defeat stress. In a second experiment, mice engaged in an operant response sequence to gain access to alcohol wherein completion of a fixed interval (FI; 5 min) schedule was reinforced with continuous access to alcohol (fixed ratio; FR1) for up to 1.8 g/kg. Alcohol-reinforced responding and subsequent alcohol consumption were assessed daily for 4 weeks prior to the 10-day social stress exposure and for 6-week post-stress. Machine learning was implemented to standardize the analysis of defeat behaviors exhibited by the intruder mouse during confrontation with an attacking resident.

RESULTS

In mice with a prior history of alcohol drinking, intermittent episodes of social defeat stress produced a significant increase in 20% EtOH consumption in preference over concurrently available water. This increased intake persisted for at least 6 weeks after the final social stress experience. Intermittently stressed mice also accelerated their anticipatory responding during the fixed interval component of the operant response chain that was reinforced by alcohol. Neither unstressed controls nor mice exposed to continuous social stress exhibited significant increases in alcohol consumption and alcohol reinforcement.

DISCUSSION

Episodic social defeat stress promotes the seeking and consumption of alcohol, extending earlier work to alcohol-experienced mice. We hypothesize that intermittent access to alcohol and intermittent episodes of social stress are additive and share common sensitizing neural mechanisms that engender excessive alcohol consumption.

To fight or not to fight: activation of the mPFC during decision to engage in aggressive behavior after ethanol consumption in a novel murine model

RATIONALE

Alcohol consumption is a common antecedent of aggressive behavior. The effects of alcohol on the decision to engage in aggression in preference over pro-social interaction are hypothesized to arise from augmented function within the medial prefrontal cortex (mPFC).

OBJECTIVE

In a newly developed procedure, we studied social decision-making in male C57BL/6 J mice based on preferentially seeking access to either sociosexual interactions with a female partner or the opportunity to attack an intruder male. While deciding to engage in aggressive vs. sociosexual behavior, corresponding neural activation was assessed via c-Fos immunoreactivity in cortical, amygdaloid and tegmental regions of interest. A further objective was to investigate how self-administered alcohol impacted social choice.

METHODS

During repeated confrontations with an intruder male in their home cage, experimental mice engaged in species-specific sequence of pursuit, threat, and attack behavior within < 2 min. Mice were then conditioned to respond at one of two separate illuminated operanda in an experimental chamber (octagon) attached to their home cage; completion of 10 responses (fixed ratio 10; FR10) was reinforced by access to either a female or a male intruder which were presented in the resident's home cage. Brains were harvested following choice between the concurrently available aggressive and sociosexual options and processed for c-Fos immunoreactivity across 10 brain regions. In two separate groups, mice were trained to rapidly self-administer ethanol prior to a social choice trial in order to examine the effects of alcohol on social choice, sociosexual, aggressive acts and postures, and concurrent c-Fos activity in the mPFC and limbic regions.

RESULTS AND DISCUSSION

Eight out of 65 mice consistently chose to engage in aggressive behavior in preference to sociosexual contact with a female when each outcome was concurrently available. Self-administered alcohol (experiment 1: 1.2 ± 0.02 g/kg; experiment 2: 0, 1.0, 1.5, and 1.8 g/kg) increased responding for the aggressive option in mice that previously opted predominantly for access to sociosexual interactions with the female. When choosing the aggressive, but not the sociosexual option, the prelimbic area of the mPFC revealed increased c-Fos activity, guiding future detailed inquiry into the neural mechanisms for aggressive choice.

Toward the explainability, transparency, and universality of machine learning for behavioral classification in neuroscience

The use of rigorous ethological observation via machine learning techniques to understand brain function (computational neuroethology) is a rapidly growing approach that is poised to significantly change how behavioral neuroscience is commonly performed. With the development of open-source platforms for automated tracking and behavioral recognition, these approaches are now accessible to a wide array of neuroscientists despite variations in budget and computational experience. Importantly, this adoption has moved the field toward a common understanding of behavior and brain function through the removal of manual bias and the identification of previously unknown behavioral repertoires. Although less apparent, another consequence of this movement is the introduction of analytical tools that increase the explainabilty, transparency, and universality of the machine-based behavioral classifications both within and between research groups. Here, we focus on three main applications of such machine model explainabilty tools and metrics in the drive toward behavioral (i) standardization, (ii) specialization, and (iii) explainability. We provide a perspective on the use of explainability tools in computational neuroethology, and detail why this is a necessary next step in the expansion of the field. Specifically, as a possible solution in behavioral neuroscience, we propose the use of Shapley values via Shapley Additive Explanations (SHAP) as a diagnostic resource toward explainability of human annotation, as well as supervised and unsupervised behavioral machine learning analysis.

Automated procedure to assess pup retrieval in laboratory mice

All mammalian mothers form some sort of caring bond with their infants that is crucial to the development of their offspring. The Pup Retrieval Test (PRT) is the leading procedure to assess pup-directed maternal care in laboratory rodents, used in a wide range of basic and preclinical research applications. Most PRT protocols require manual scoring, which is prone to bias and spatial and temporal inaccuracies. This study proposes a novel procedure using machine learning algorithms to enable reliable assessment of PRT performance. Automated tracking of a dam and one pup was established in DeepLabCut and was combined with automated behavioral classification of "maternal approach", "carrying" and "digging" in Simple Behavioral Analysis (SimBA). Our automated procedure estimated retrieval success with an accuracy of 86.7%, whereas accuracies of "approach", "carry" and "digging" were estimated at respectively 99.3%, 98.6% and 85.0%. We provide an open-source, step-by-step protocol for automated PRT assessment, which aims to increase reproducibility and reliability, and can be easily shared and distributed.

Neurobiological Bases of Alcohol Consumption After Social Stress

The urge to seek and consume excessive alcohol is intensified by prior experiences with social stress, and this cascade can be modeled under systematically controlled laboratory conditions in rodents and non-human primates. Adaptive coping with intermittent episodes of social defeat stress often transitions to maladaptive responses to traumatic continuous stress, and alcohol consumption may become part of coping responses. At the circuit level, the neural pathways subserving stress coping intersect with those for alcohol consumption. Increasingly discrete regions and connections within the prefrontal cortex, the ventral and dorsal striatum, thalamic and hypothalamic nuclei, tegmental areas as well as brain stem structures begin to be identified as critical for reacting to and coping with social stress while seeking and consuming alcohol. Several candidate molecules that modulate signals within these neural connections have been targeted in order to reduce excessive drinking and relapse. In spite of some early clinical failures, neuropeptides such as CRF, opioids, or oxytocin continue to be examined for their role in attenuating stress-escalated drinking. Recent work has focused on neural sites of action for peptides and steroids, most likely in neuroinflammatory processes as a result of interactive effects of episodic social stress and excessive alcohol seeking and drinking.

Neural circuit mechanisms that govern inter-male attack in mice

Individuals of many species fight with conspecifics to gain access to or defend critical resources essential for survival and reproduction. Such intraspecific fighting is evolutionarily selected for in a species-, sex-, and environment-dependent manner when the value of resources secured exceeds the cost of fighting. One such example is males fighting for chances to mate with females. Recent advances in new tools open up ways to dissect the detailed neural circuit mechanisms that govern intraspecific, particularly inter-male, aggression in the model organism Mus musculus (house mouse). By targeting and functional manipulating genetically defined populations of neurons and their projections, these studies reveal a core neural circuit that controls the display of reactive male-male attacks in mice, from sensory detection to decision making and action selection. Here, we summarize these critical results. We then describe various modulatory inputs that route into the core circuit to afford state-dependent and top-down modulation of inter-male attacks. While reviewing these exciting developments, we note that how the inter-male attack circuit converges or diverges with neural circuits that mediate other forms of social interactions remain not fully understood. Finally, we emphasize the importance of combining circuit, pharmacological, and genetic analysis when studying the neural control of aggression in the future.

Once bitten, twice shy: Aggressive and defeated pigs begin agonistic encounters with more negative emotions

Aggression between unfamiliar commercial pigs is common and likely invokes strong emotions in contestants. Furthermore, contest outcomes affect subsequent aggressive behaviour, suggesting a potential lasting influence on affective state. Here we used a combination of qualitative and quantitative methods to assess the emotional expression of pigs in agonistic encounters. We investigated how recent victory or defeat influences emotions expressed in a subsequent contest, and the role of aggressiveness as a personality trait in emotional expression. We observed the pre-escalation contest behaviour (second contest; age 13 wks) in animals of different aggressiveness (categorised using two resident intruder tests as Agg+ or Agg-, age 9 wks), which had recently won or lost a contest (first contest; 10 wks). We measured gaze direction and ear position. Observers watched video clips of the initial 30 s of the second contest and evaluated the emotional expression of 57 pigs (25 contest 1 winners, 32 contest 1 losers) using qualitative behavioural assessment (QBA) with a fixed list of 20 descriptive terms. QBA identified three principal components (PCs), accounting for 68% of the variation: PC1 (agitated/tense to relaxed/content), PC2 (fearful/aimless to confident/enjoying) and PC3 (listless/ indifferent). Agg- pigs and males showed a more positive emotionality (PC2). PC1 and PC3 were unaffected by first contest outcome and aggressiveness. Agg+ pigs were more likely to hold their ears back (X2 =7.8, p = 0.005) during the early contest period. Differences in attention were detected in the contest outcome × aggressiveness interaction (χ24.3, p = 0.04), whereby approaching the opponent was influenced by winning and losing in the Agg- pigs only. QBA and gaze behaviour reveal differences in emotional valence between pigs of different aggressiveness: less aggressive pigs may be more susceptible to the emotional impact of victory and defeat but overall, more aggressive pigs express more negative emotionality at the start of agonistic encounters.

Quantitative standardization of resident mouse behavior for studies of aggression and social defeat

Territorial reactive aggression in mice is used to study the biology of aggression-related behavior and is also a critical component of procedures used to study mood disorders, such as chronic social defeat stress. However, quantifying mouse aggression in a systematic, representative, and easily adoptable way that allows direct comparison between cohorts within or between studies remains a challenge. Here, we propose a structural equation modeling approach to quantify aggression observed during the resident-intruder procedure. Using data for 658 sexually experienced CD-1 male mice generated by three research groups across three institutions over a 10-year period, we developed a higher-order confirmatory factor model wherein the combined contributions of latency to the first attack, number of attack bouts, and average attack duration on each trial day (easily observable metrics that require no specialized equipment) are used to quantify individual differences in aggression. We call our final model the Mouse Aggression Detector (MAD) model. Correlation analyses between MAD model factors estimated from multiple large datasets demonstrate generalizability of this measurement approach, and we further establish the stability of aggression scores across time within cohorts and demonstrate the utility of MAD for selecting aggressors which will generate a susceptible phenotype in social defeat experiments. Thus, this novel aggression scoring technique offers a systematic, high-throughput approach for aggressor selection in chronic social defeat stress studies and a more consistent and accurate study of mouse aggression itself.

Expressions of emotions across species

What are emotions and how should we study them? These questions give rise to ongoing controversy amongst scientists in the fields of neuroscience, psychology and philosophy, and have resulted in different views on emotions [1-6]. In this review, we define emotions as functional states that bear essential roles in promoting survival and thus have emerged through evolution. Emotions trigger behavioral, somatic, hormonal, and neurochemical reactions, referred to as expressions of emotion. We discuss recent studies on emotion expression across species and highlight emerging common principles. We argue that detailed and multidimensional analyses of emotion expressions are key to develop biology-based definitions of emotions and to reveal their neuronal underpinnings.