The scent of emotions: A systematic review of human intra‐ and interspecific chemical communication of emotions

Abstract Objective The sense of olfaction has been considered of minor importance in human communication. In recent years, evidence has emerged that humans might be influenced by unconscious messages sent through chemosignals in body odors. Data concerning the ability of humans to recognize fear, maybe related to the evolutionary role of these emotions in the fight‐or‐flight reactions, are well known. Methods To further understand the role of emotional chemosignals in mediating communication in humans and its influence on animal behaviors, we conducted a systematic literature review. Results Chemosignals derived from axillary odors collected under a variety of emotional stimuli and sad tears in humans affect receivers' social interactions, danger detection and risk‐taking behavior, social aspects of eating, and performance under stressing conditions. In addition, beyond the fight‐or‐flight response, even the body odors of happiness can be perceived by others. Furthermore, human chemosignals can influence behaviors and stressful responses in animals, particularly dogs and horses, which may partially explain their special relationship with humans. Conclusion Our review highlights the importance of chemosignaling in human intra‐ and interspecific interactions and suggests the need for further investigations, both in physiological conditions and in patients with psychiatric or neurodegenerative disorders.


2014
). This observation drives scientific efforts toward the study of the other senses, leaving the olfactory function largely unexplored.
Nevertheless, it is well known that primates maintain a variety of sebaceous and apocrine skin glands (Montagna & Yun, 1972) as well as an excellent olfactory sensibility expressed as ability in discriminating odorants involved in reproductive signaling, even if compared to dogs and rats (Laska, 2000). In addition, a number of studies showed in primates the involvement of olfaction, not only in scent marking (Heymann, 2006), but also in social and sexual behaviors (Kappeler, 1998), the communication of reproductive status or the pair-bonding (Snowdon, Ziegler, Schultz-Darken, & Ferris, 2006).
Almost 400 intact OR genes have been identified in humans, a small number in comparison with dogs and rodents. Once odorant molecules bind to ORs, the signal transduction is mediated by the cilia of olfactory sensory neurons (OSNs) through the increase in intracellular cyclic adenosine monophosphate leading to neuron depolarization. OSNs converge onto glomerular structures in the olfactory bulb from which mitral cells project directly to the primary cortex, without thalamic relay, thus distinguishing the sense of olfaction from all the other human senses (Menini, 2010). Nevertheless, in recent years the involvement of the medio-dorsal nucleus of the thalamus (MDT) in processing olfactory stimuli has been postulated (Price and Powell, s.d.) as the MDT receives inputs from all the primary olfactory areas including the piriform cortex and some secondary olfactory areas, potentially involved in olfactory stimuli processing including odor identification, discrimination, attention, and learning (Courtiol & Wilson, 2015). The detection of pheromones in humans was thought to be completely segregated by the MOS and mediated by the VNO, although its functional involvement and presence is still questioned in humans (Meredith, 2001). The VNO is a tubular structure situated in the nasal septum, part of the accessory olfactory system and specialized in detecting pheromonal involatile signals through direct physical contact (Bhatnagar & Smith, 2001). The accessory olfactory bulb, receiving inputs from vomeronasal sensory neurons axons, projects mainly to the medial and posteromedial cortical amygdala, and then to the hypothalamus, controlling reproductive and social behavior (von Campenhausen & Mori, 2000).
Nevertheless, the AOS and MOS functions are more integrated than previously thought, as both structures can respond to the same chemical stimuli and both sensory systems send projections to brain areas that are involved in mediating pheromonal responses (Brennan & Zufall, 2006).
Olfactory communication is of pivotal importance in animals' social interaction. Body odors and volatile compounds in urine, feces, or blood have been demonstrated to be a warning signal to prey species (Schauber, 2008), activating many autonomic, endocrine, and behavioral responses (Ulrich-Lai & Herman, 2009). For example, mice smelling a mixture of pyrazine from the wolves' urine increased both vigilance behaviors and activity of the neurons in the AOS; the same substances suppress the approach of deer to feeding areas while eliciting fear responses (Osada, Miyazono, & Kashiwayanagi, 2015). Some authors hypothesized that predator odors could be detected by specific olfactory structure as MOS-mediating responses to volatile cues (Firestein, 2001) and AOS for chemical cues or pheromones (Breer, Fleischer, & Strotmann, 2006). Specific brain areas as amygdala and hippocampus play a key role in activating autonomic and endocrinological responses (e.g., hypothalamic-pituitary-adrenal axis). Amygdala is also involved in the unconditioned fear behavior related to predator odor and in the retrieval of contextual fear memory associated with prior predator odor experiences.
It is widely recognized that humans' five senses work together in providing information and that signals received from one sense can modulate the information received from another in a multisensory way (Stein & Meredith, 1993). The relationship between visual, auditory, and somatosensory inputs, the so-called "physical senses," has been largely studied (Alais, Newell, & Mamassian, 2010). With regard to olfaction, we know that interaction with taste is fundamental in appetite modulation and perceptions of the foods (McCrickerd & Forde, 2016). Moreover, visual perception can affect olfactory identification (i.e., in white versus red wine identification by expert tasters as demonstrated by the study of Morrot, Brochet, & Dubourdieu (2001) and vice versa, modulating food-images attractiveness, human faces pleasantness (Cook et al., 2015;Luisa Demattè, Sanabria, & Spence, 2006) or facial emotion recognition (Seubert, Gregory, Chamberland, Dessirier, & Lundström, 2014).

Summations
• Humans are able to sense and react to intraspecific chemosignals enclosed in body odors, but the exact composition of chemosignals is unknown and data on transmission of "positive emotions" trough body odors are lacking • As data on the role of chemosignaling in demented and psychiatric patients are missing, there is high potential for further studies on emotional chemosignaling in humans • Dogs and horses are influenced by human emotional chemosignals Limitations • Our search strategy was restricted to English-language publications, published between January 1970 and April The sense of olfaction is unique in projecting directly to the amygdala and the orbitofrontal cortex, thus providing a close connection with the limbic system, expressly tasked with emotion processing (Hackländer, Janssen, & Bermeitinger, 2019;Krusemark, Novak, Gitelman, & Li, 2013).
A number of behavioral studies demonstrated that olfactory cues makes memories more emotional and evocative if compared to other sensory stimuli (Herz, 2016;Herz, Eliassen, Beland, & Souza, 2004). Moreover, functional magnetic resonance imaging (fMRI) studies demonstrated that memories elicited by odor perception activate specific neuroanatomical area if compared to other sensory stimuli (Herz et al., 2004).
Olfaction is also involved in odor disease avoidance: The inflammatory process leads to the release of volatile molecules in urine and feces that are recognized by conspecifics, providing information about the health status of the odor donors. The detection of sick individuals via odor cues is well known in animals and helps to avoid disease transmission inhibiting social interactions (Arakawa, Cruz, & Deak, 2011). In humans, disease-specific (e.g., infectious or metabolic disease) volatile organic compounds have been identified (Shirasu & Touhara, 2011). Considering the dramatic role of infections in human evolution, the ability to detect olfactory cues indicating sickness could represent an adaptive survival mechanism. Some experimental studies demonstrated an unconscious ability of healthy subjects to recognize and find repulsive body odor obtained from "sick" subjects (Olsson et al., 2014); smelling these body odors activate the odor networks as shown by fMRI (Regenbogen et al., 2017). Nevertheless, many questions remain still open and literature is lacking about the neural processes underlying the ability of humans to detect sickness.
In the last decades, it has become clear that also humans have excellent olfactory abilities (McGann, 2017). The exceptional ability of humans to discriminate a big number of odorants (Bushdid, Magnasco, Vosshall, & Keller, 2014) despite the limited number of functional ORs depends on a combinatorial receptor coding scheme (Malnic, Hirono, Sato, & Buck, 1999). Scientific interest has been centered on the role of olfactory communication in shaping social interactions through molecules produced in specific emotional states (Lübke & Pause, 2015). Such molecules mediating interindividual communicative exchanges were firstly classified as pheromones and are now named chemosignals (Doty, 2010).
The question if and how humans may react to chemosignals is, indeed, challenging and not completely answered by experimental studies. Data on intraspecific communication between different species of animals (Brennan, 2010;Wyatt, 2010Wyatt, , 2014aWyatt, , 2014b confirm the common observation that animals communicate with each other through body odors. More surprisingly, some experimental studies suggest that also humans may be influenced in their interpersonal relationships and behaviors by the unconscious messages sent through chemosignals enclosed in body odors (de Groot, Smeets, Kaldewaij, Duijndam, & Semin, 2012).
Chemosignals are molecules excreted by animals as answer to physical distress and emotions and are able to elicit behavior or physiological responses from other animals (Petrulis, 2013). Despite this definition, until now, there is no clear evidence of which molecules are able to vehicle emotions, several molecules have been indicated as chemosignals, and these molecules have to be differentiated from odors and volatile substances (Table 1 and 2). Among these molecules, the testosterone metabolite androstadienone has been indicated as a putative chemosignal and suggested to be able to communicate dominance and social threat by several studies (Banner, Frumin, & Shamay-Tsoory, 2018;Frey, Weyers, Pauli, & Mühlberger, 2012;Hornung, Kogler, Wolpert, Freiherr, & Derntl, 2017;Zhou et al., 2014).
In addition, research on chemosignaling is focusing on the transmission of emotional states.
Preliminary studies investigated the involvement of chemosignals in conveying emotional states from "a sender" to "a receiver." In 2000, Chen and Haviland-Jones were able to demonstrate for the first time that human subjects can recognize the emotion of another human subject by sniffing odors collected by axillary pads (Chen & Haviland-Jones, 2000). In the following years, a number of further evidences confirmed that human body odors vary according to emotional states of the donors and that these changes can be perceived by receivers (Pause, 2004a;Pause, Adolph, Prehn-Kristensen, & Ferstl, 2009;Prehn, Ohrt, Sojka, Ferstl, & Pause, 2006).
The majority of research on communication via human body odors has focused on the transmission of the so-called "negative emotions" (i.e., fear, stress or anxiety; de Groot & Smeets, 2017), based on the evolutionary significance of potential activation of adrenergic-mediated stress response system. In subsequent studies, similar results have been obtained with "positive emotions" as happiness or sexual arousal (Iversen, Ptito, Møller, & Kupers, 2015;Zhou, Hou, Zhou, & Chen, 2011) showing the complexity of chemosignaling in human's communication.
Neurofibrillary tangles early accumulate in the key areas for olfactory function in AD (Kovács, Cairns, & Lantos, 1999;Ohm & Braak, 1987), and neuroimaging studies demonstrate atrophy in the primary olfactory cortex and hippocampus in AD patients (Kotecha et al., 2018;Vasavada et al., 2015). Interestingly, impaired ability to identify different odors seems to predict the progression of cognitive decline in subjects with mild cognitive impairment (Devanand et al., 2000). Limited evidences suggested that olfactory dysfunction might be useful to differentiate AD from another type of dementia (Park, Lee, Lee, & Kim, 2018).
Also in Parkinson's disease, the olfactory dysfunction plays a key role in the diagnosis, as its evaluation is included in the diagnostic course, in particular in distinguishing Parkinson's disease from other parkinsonian syndromes (Suchowersky et al., 2006). In Parkinson's disease, olfactory impairment appears years before the clinical manifestation of the disease, remains stable over time, and affects more than 90% of patients (Doty, 2012). Moreover, in longitudinal studies olfactory impairment can predict the rate of evolution toward dementia (Baba et al., 2012).
Recent data suggest that humans' chemosignals could also be per- Nevertheless, many questions remain unanswered: Little is known about the brain areas involved in the recognition of the emotions transmitted through chemosignals, as well as the consequences of neurodegenerative or psychiatric pathologies on the ability to recognize the chemical messages. Furthermore, whether chemosignals are recognized through the primary olfactory system or through the VNO in humans remains controversial (D'Aniello, Semin, Scandurra, & Pinelli, 2017;Meredith, 2001) and the identification of active compounds involved in chemosignaling is far from completion. As geriatricians, we are particularly interested in understanding the different reactions of cognitive impaired patients to their professional and familiar caregivers' chemosignals (Rippon et al., 2019).
Here, we systematically review the studies on the communication of emotions by chemosignals in humans and between humans and other species. The understanding of emotional communication through chemosignals will increase our understanding of intraspecific and interspecific communications.

| Eligibility criteria
Inclusion criteria were based on the Participants, Intervention, Comparator, Outcomes, and Study design, the PICO model was built as follows: Participants: We included studies investigating the effects of human-derived emotional chemosignals on human and animal receivers.

Interventions:
We included only studies analyzing the responses to emotional stimuli derived by body odors collected from a sender under an emotional condition. Studies with synthetic substances or hormonal stimuli were excluded.
Comparator: A control stimulus had to be presented to the receiver and included body odors obtained during exercise or after a neutral stimulus, unused sweat pads, or saline solutions.
Outcomes: We included studies investigating the ability of an emotional body odor to elicit the same emotion in the sender as compared to a control stimulus. Measures could be fMRI, facial electromyography (EMG), skin conductance response (SCR), electroencephalography (EEG), cardiac activity or cognitive, affective, behavioral, or perceptual tasks.

Study design:
We included English-language and peer-reviewed studies with no limitations due to study type or publication date.

| Information source
This systematic review was performed according to the Preferred Reporting Items for Systematic Reviews and Meta-analyses (PRISMA) checklist from January 1970 to April 2019.
The search strategy was conducted to find relevant studies from the MEDLINE, EMBASE, Cochrane Library, and PsychINFO databases.

Odor Volatile molecule Pheromone
Blend of different moieties released in organic fluids that varies according to species, sex, age, genotype, and endocrine state and/or the property of certain substances, in very small concentrations, to stimulate chemical sense receptors.
Chemical that has a high vapor pressure at ordinary room temperature.
A chemical released by one organism that modulates the behavior or physiology of a second organism of the same species, which ranges from small, volatile molecules, and sulfated steroids to large families of proteins. Its principal properties are as follows: • The synthesized molecule/combination of molecules should elicit the same response as the natural stimulus in the bioassay. • It should act in this way at natural concentrations. At high concentrations, spurious results may occur as nonpheromones may stimulate receptors; • For multicomponent pheromones, experiments should demonstrate that all compounds in the combination are necessary and sufficient to elicit the full response; • Only this molecule or the proposed combination of molecules elicits the effect (unlike other similar molecules or combinations that the animal would normally encounter); • There should be a credible pathway for the pheromone signal to have evolved by direct or kin selection. In evolutionary terms, to be a signal, both the emission and reception of the pheromone signal should have evolved for a particular function.
A manual search of these articles' reference lists was performed to capture additional articles for consideration; this search allowed us to find one article from Kamiloğlu, Smeets, de Groot, and Semin (2018).

| Search strategy
The search evaluated articles using the search terms:

| Study selection
Two experienced reviewers (EC and UQ) identified all studies meeting the inclusion criteria to be included for the full review. Each reviewer independently selected studies for inclusion in the review, and discrepancies were resolved by mutual consensus.

| Data extraction and analysis
This search query returned 451 (PubMed) + 692 (EMBASE) + 11 (PsychINFO) + 74 (Cochrane) articles for review. After removing duplicates, we excluded 741 articles (Figure 1). Fifty-seven articles were reviewed in full text by the authors and considered for evaluation.
Selected articles for review were published between 2000 and 2018.
We were able to find on the Web two relevant studies as unpublished dissertation; however, we decided to exclude those studies from this review as they were not peer-reviewed (Hatcher, s.d.; Owen, s.d.).
Twelve articles were excluded after reading the full text as they were considered nonpertinent. Based on the full-text review, forty-five articles were selected for full-text, in-depth review (Table 3). A flow diagram of the selection procedure is included in Figure 1.
The following variables were extracted from each study: year of publication, chemosignal type, emotion induction, odor control condition, assessment of induced emotion, male/female senders and receivers, olfactory function assessment, stimuli collection material, stimuli presentation, main outcome.
Data were collected using Microsoft Excel (version 16.11).
This study does not contain any studies with human participants or animals performed by any of the authors. For this type of study, formal consent is not required.

| RE SULTS
The studies analyzed were highly heterogeneous in methodology: They differed in the stimulus chosen (sweat or tears); in the method used for the induction of emotional response in the donors (ranging between watching different kinds of videos, to extreme sports experience); in the kind of emotion evaluated; in the subjects enrolled as donors or receivers, differences in subjects included age, sex, and sexual orientation; in the main outcomes and the methods of measurement. Table 3 describes the key characteristics of the studies included in this review.

| Intraspecific communication
Forty-two studies investigated intraspecies chemosignals communication in humans. Among these, in 40 studies chemosignals derived from axillary sweat extracts from a total of 568 male and 327 female donors; in the remaining two studies, chemosignals derived from sad tears from a total of 6 female donors (Gelstein et al., 2011;Oh, Kim, Park, & Cho, 2012). All donors were healthy adults (minimum and maximum age of 18 and 50 years, respectively).
In one article, donors were partners of female receivers .
In one study, control body odors were collected during an emotionally neutral situation (attending a regular class; Rocha et al., 2018).
Main outcomes were very heterogeneous too: Correct identification of the target emotion or odor rating was the main outcome of five studies (Ackerl et al., 2002;Chen & Haviland-Jones, 2000;Haviland-Jones et al., 2016;Iversen et al., 2015;. The influence of emotional chemosignals on cognitive tasks like performing word association while smelling one of the three types of olfactory stimuli was used by one research group (Chen, 2006). Priming of facial affect perception was the main outcome in one study (Pause, 2004a). Recognition of facial expressions after the exposition to anxiety or relaxed body odors was the main outcome in 4 papers (Mutic et al., 2016;Rocha et al., 2018;Zernecke et al., 2011;Zhou & Chen, 2009). The amplitude of the startle reflex recorded in the context of chemosensory anxiety signals was the main outcome in 4 studies (Adolph et al., 2013;Lübke et al., 2017;Pause et al., 2009;Prehn et al., 2006).
Amygdala activation during an fMRI session and ability to recognize ambiguous facial expression in relation to exposure to emotional stress body odors was used in one paper (Mujica-Parodi et al., 2009). Brain areas activation after administration of chemosensory stimuli (Gelstein et al., 2011;Mutic et al., 2017;Prehn-Kristensen et al., 2009;Radulescu & Mujica-Parodi, 2013;Wintermann et al., 2013;Wudarczyk et al., 2015Wudarczyk et al., , 2016Zheng et al., 2018; as main outcome was analyzed in 9 studies. Haegler et al. investigated the risk-taking behavior in computerized card games after smelling anxiety body odor (Haegler et al., 2010). Adolph et al. (2010) measured as main outcome skin conductance response of receivers in response to competition sweat. Authors investigated the influence of anxiety body odor on chemosensory event-related potentials recorded during an EEG session in three studies (Adolph et al., 2013;Pause et al., 2010;Rubin et al., 2012). Measure of anxiety through the Spielberger's state-trait anxiety inventory was evaluated in one study . In seven studies, authors investigated the ability to reproduce the same facial-muscle configuration of the sender in the receiver with EMG (de Groot et al., 2012(de Groot et al., , 2014a(de Groot et al., , 2014bGroot, Smeets, Rowson, et al., 2015;Kamiloğlu et al., 2018). Singh et al. (2018) analyzed the effect of anxiety signals on the performance of dentistry students on three different dental procedures. Dalton and colleagues evaluated the influence of psychosocial stress body odor on social judgment (rating warmth and competence about women depicted in video scenario) (Dalton et al., 2013).
Appetite assessment by a visual analog scale (VAS) and food intake in men exposed to the smell of sad tears or trickled saline was the main outcome in 1 study (Oh et al., 2012). Cardiac parasympathetic activity measured in receivers was the main outcome in 1 case (Ferreira et al., 2018).

| Interspecific communication
We found only three studies investigating the ability of animals to react to human chemosignals.
In 2016, for the first time in literature, Siniscalchi et al. tested the ability of 31 domestic dogs of various breeds (11 males and 20 females) to react to human chemosignals (Siniscalchi et al., 2016). System reactions of 7 male horses in response to exposure to human happy and fearful chemosignals (Lanata et al., 2018). The main outcome was time-frequency analysis of horses' heart rate variability.

| D ISCUSS I ON
The understanding of communication beyond words and body language is taking great interest; chemosignals transmitted through body odors may play a role in the communications in humans and between humans and other species.
The first peer-reviewed article on this topic was published in 2000 by Chen and Haviland-Jones (2000): The authors demonstrated that women performed better at olfactory identification of emotions than men, confirming previous data showing a better ability of women to recognize visual and auditory emotional signals (Brody & Hall, 2008).
Further studies confirm that women are better receivers for chemosignals than men (de Groot et al., 2014a); hence, the majority of the studies involves women as receivers and male as donors. It is clear that chemosignals from donors of the opposite sex are more effective than those from the same sex (Martins et al., 2005) pointing out that chemosignals may be important for reproductive purposes.
On the other hand, there does not seem to be a different perception of chemosignals between different ethnicities, suggesting that chemosignaling communication could act beyond ethno-cultural boundaries .
A study on sexual appealing showed reduced physiological measures of arousal and lower levels of testosterone in men who sniffed tears from sad women compared to a control (Gelstein et al., 2011).
Moreover, a study on the ability to react to body odors from partners demonstrated that intimacy enhances the detection of emotional cues, although not consciously . Receivers are generally unable to consciously recognize the stimulus and name the body odor. On the other hand, this is not surprising, as olfaction has been termed "the mute sense" (Ackerman, 1991).
It has been suggested that increased perception and reaction to anxiety and fear may be responsible for social anxiety; in fact, Pause et al. demonstrated that the defense reflex and the required neuronal resources of anxiety-related chemosignals were enhanced as in socially anxious receivers as compared to nonsocially anxious ones Pause et al., 2010).
Overall, negative emotions of the donor, as anxiety and fear, seem to be perceived by and influence social behavior in the recipi- In cognitively healthy subjects, anxiety chemosignals may influence job performances as it has been demonstrated by Singh et al.: In their experiment, authors showed that dentistry students worsened their professional performances if exposed to body odors produced in an anxiety-inducing situation (Singh et al., 2018).
Notably in the majority of studies, the detection rate of the target emotion was very poor, suggesting that chemosignaling communication in humans acts below awareness Zhou & Chen, 2009.  et al., 2016). Moreover, dogs exposed to human happiness chemosignals appeared more confident with strangers, implying that a relaxed mood of owners calms their pet dogs . In horses, human fear and happiness chemosignals induced sympathetic and parasympathetic changes indicating emotional activation (Lanata et al., 2018). However, this latter study, while providing interesting data, remains preliminary, due to the little sample size.
Overall fear, anxiety, dominance, and sexual arousal are the most recognized emotions through chemosignals (de Groot & Smeets, 2017), whereas the demonstration of recognition of happiness is less frequent (Groot, Smeets, Rowson, et al., 2015). This was also true in humans if the pattern of emotional recognition used is vi-  Sander, & Vuilleumier, 2004), which make the data less robust and awaiting confirmation. Alternatively, it is possible that emotions such as fear, anxiety, dominance, and sexual arousal could be more easily recognized in contrast to happiness, due to their major evolutionary relevance and reproductive role.

| CON CLUS IONS
Despite the wide heterogeneity between studies and the small sample sizes analyzed, the evidences highlight the importance of chemosignals in social interaction, empathy with the partner, social judgment, danger detection, social aspect of eating, risk-taking behavior, stressful performance, and perhaps perception of happiness.
Less evidence of a role of chemosignals in personality disorders and psychiatric pathologies is available, and there are no data on chemosignaling neurodegenerative and age-related brain diseases.
Improving our knowledge on chemosignal communication in patients with psychiatric or neurodegenerative disorders could be of paramount importance to better understand the disease pathophysiology and to develop new diagnostic and therapeutic strategies, and to this extent, the adoption of a clear evidence-based study design is of fundamental importance.

ACK N OWLED G M ENT
This research received no specific grant from any funding agency in the public, commercial, or not-for-profit sectors.

CO N FLI C T O F I NTE R E S T
The authors have no conflicts of interest to declare.

AUTH O R CO NTR I B UTI O N
PDA and BDA conceived and supervised the study, AS and MM supervised the study, EC and UQ retrieved the data, and all the authors wrote and approved the manuscript.

DATA AVA I L A B I L I T Y S TAT E M E N T
Data sharing is not applicable to this article as no new data were created or analyzed in this study.