Emotions in Animals: The end of a Debate?

mechanisms (internal push effects) and social norms (external pull effects) would influence the encoding and the decoding of the vocal expression of emotions in human and other animals (Grandjean et al., 2006; Scherer, 1988, 2009c; see Figure 9).

Figure 9: Adaptation of Brunswik’s lens model, including the influences of conventions, norms and display rules - pull effects - and psychobiological mechanisms - push effects - on the encoding of emotional vocalizations produced by the speaker and the decoding made by the listener on reciprocal influences of these two aspects on attributions. An emotional content is thus expressed by distal indicators cues (e.g. acoustic features) that are perceived by a listener who on the basis of proximal percept and contextual information makes a subjective attribution of the speaker’s emotional states (Grandjean et al., 2006).

Affective sciences through the main theories of emotions, recognized the crucial role of evolutionary processing such as natural selection in the emergence of emotions. For instance, despite the current debate, the universality of basic emotions suggests inherited mechanisms based on human evolution. If this hypothesis of inheritance is correct, similar processes should be found in other species as well, in particular in non-human primates (NHP), our closest relatives.

2.1.2.2. Emotions in Animals: The end of a Debate?

Based on René Descartes’ words, “Animals are like robots: they cannot reason or feel pain.”

(Proctor et al., 2013, p. 883). Hopefully, in the 21^st century, this Cartesian point of view is no longer shared in affective sciences. However, the necessity to define or even study emotions

are still debated in animal research (de Vere & Kuczaj, 2016; de Waal, 2011; LeDoux, 2012).

Many reasons may explain this ongoing debate. First, our anthropocentric view of human emotions conducted some scientists to not recognize complex emotional processes in animals or to consider them differently because they would necessarily be less important than the ones experienced by humans (Proctor et al., 2013).

Second, the “Anthropo-denial” inherited in research from the avoidance of anthropomorphism to explain animals’ behaviours (de Waal, 2018; Panksepp, 2011a).

Anthropomorphism, indeed, cannot be used directly to explain results; however, it should be considered in the formulation of hypotheses to deal with certain types of primary psychobiological processes that we might share with other animals (Panksepp & Burgdorf, 2003) .

Third, the ongoing question of existing innate circuitries in the human brain (LeDoux, 2012). Do emotions exist by nature or are they specific to the human mind? (Charland, 2002;

Griffiths, 2004).

Finally, the lack of consensus in the use of the terms emotion, affect or feeling. Usually, in psychology the notion of affect encompassed several classes or categories of mental states such as emotions, moods or attitudes (Frijda & Scherer, 2009). The notion of feeling refers to the subjective perception of emotional states and their accompanying somatic responses.

For now, feelings can only be assessed by verbal reports (Anderson & Adolphs, 2014). Yet, in the animal literature, the terms emotion, feeling and affect are often used as synonyms (Anderson & Adolphs, 2014; Boissy et al., 2007; de Vere & Kuczaj, 2016). Hence, it has become crucial in affective sciences, neurosciences, primatology or even ethology to find an agreement on the mechanisms that define animal emotions. From this, we will use in this section and the following chapters, the more general term of affect when referring to animals’ emotional experiences (humans excluded).

Despites all the debates, researchers such as Jaak Panksepp, have greatly contributed to the exploration of affective mechanisms in animals. In fact, pioneer in the field of comparative affective neurosciences, he often described seven different systems in the mammalian brains that promote affective actions and associated feelings: i) SEEKING for expectancy; ii) RAGE for anger; iii) FEAR for anxiety; iv) LUST for sexuality; v) CARE for nurturance; vi) PANIC for separation; and vii) PLAY for joy (Panksepp, 2010, 2011a, 2011b; see Figure 10).

Those evolutionary and primal affects are the components of the basic affective circuits of

mammalian brains, involving sub-cortical regions in the limbic system (amygdala, hypothalamus, cingulate, hippocampus) and cortical areas such the frontal or temporal lobes (Montag & Panksepp, 2017; Panksepp, 2011b).

Figure 10: The four major affective operating systems. They are defined primarily by genetically coded but experientially refined neural circuits that generate well-organized behaviour sequences (Panksepp, 2011b).

Going further, Panksepp also investigated the affective consciousness in animals. For this purpose, he defined the term consciousness as “The brain states that have an experiential feel to them, and it is envisioned as a multi-tiered process that needs to be viewed in evolutionary terms, with multiple layers of emergence” (Panksepp, 2005, p.32). Then, he distinguished: i) the primary-process consciousness, reflecting sensory – perceptual feelings and emotional – motivational experiences; ii) the secondary-consciousness, referring to the capacity to have thoughts about external events; and iii) the tertiary form of consciousness, related to awareness, the ability to remember experiences and to transform simple thoughts in linguistic and symbolic meanings (Panksepp, 2005).

Affective consciousness in animals is still strongly discussed in research (Boissy et al., 2007).

Yet, this clear distinction between different levels of consciousness could allow scientists to consider the existence of the primary and secondary processes in animals. These two first levels could be the primitive precursors of more complex layers of consciousness in our evolutionary history. Thus, the third process, more complex, would be specific to the human mind. Therefore, animals could be conscious at a certain level about their affective

experiences but they should be not consciously aware about them (Berridge, 2002;

Panksepp, 2005, 2010). This lack of tertiary consciousness would explain why only humans can stimulate or regulate their emotions through rumination or self-criticism for example (Gilbert, 2015). However, it is also possible that animals have these abilities as well, suggesting that the available scientific methodologies are not able to assess them.

Researchers may not be able to penetrate these kinds of thoughts (Panksepp, 2014).

According to Arnellos and Keijzer, the term animal refer to “Collections of cells that are connected and integrated to such an extent that these collectives act as unitary, large free-moving entities capable of sensing macroscopic properties and events” (Arnellos & Keijzer, 2019, p.1).

This definition encompasses a very large number of species, independently of their phylogenetic proximity to humans.

The ability to recognize and respond appropriately to affective expressions of others has biological fitness benefits for both the signaller and the receiver (Schmidt & Cohn, 2001).

From this evolutionary point of view, the capacity to express and identify affective cues, in conspecifics or even in heterospecifics, should be found across a large number of species:

“The assumption that the other animals are unfeeling behavioural zombies seems evolutionarily improbable” (Panksepp & Burgdorf, 2003, p. 536).

In fact, the literature often describes the ability of domestic dog to recognize affective cues associated with pictures or/and vocalizations. For instance, in a cross-modal looking paradigm, Albuquerque and collaborators demonstrated in seventeen dogs their abilities to look significantly longer at the pictures representing faces (dogs or humans) that were congruent with the valence (positive or negative) of the vocalizations. These results underlined first, the correct discrimination of domestic dogs between positive and negative affects in both humans and dogs; second, the importance of affective recognition in heterospecifics for such species living in mixed groups and socially interacting with humans (Albuquerque et al., 2016).

In a famous and controversial study of 1999, Panksepp and Burgdorf revealed that laughter was not only shared between primates but was also present in very distant species such as rats (Rattus norvegicus). The “Laughing rats” were indeed able to vocalize laughter-like responses in context of play with conspecifics and heterospecifics (experimenters).

Interestingly, the authors hypothesized that rats’ laughter would correspond to the ones expressed by humans in early childhood (Panksepp & Burgdorf, 2003).

Research on affective experiences in animals is often accused of mammalcentrism (Proctor et al., 2013). However, molluscs or even insects also seem to have a certain type of primitive affects, making the evolutionary explanation only one factor to understand emotional processing in the human mind (Bolhuis & Wynne, 2009). For example, Anderson and Adolphs showed the potential existence of affective behaviours in octopus. In fact, the switch between camouflage (freezing behaviour) and ink expulsion – propulsion (avoidance behaviour) of an octopus when confronted to a threatening situation (e.g. predator) would demonstrate the gradation of affective intensity in the mollusc. Furthermore, the same authors as well as LeDoux, described neural circuitries associated to specific postures in flies (drosophila; see Figure 11) and approach-avoidance behaviours in worms (c.elegans), emphasizing some primitive aspects of affective experiences in phylogenetically very distant species (Anderson & Adolphs, 2014; LeDoux, 2012).

Figure 11: Illustration of the antithesis theory of Darwin in which opposite affective experiences produce behaviourally opposite expressions. a) Human sadness (left image) and human happiness (right image); b) Antithetical postures in dogs (from Darwin, 1872); c) opposite postures in flies (Anderson & Adolphs, 2014).

Nevertheless, to improve our knowledge about the evolutionary history of human emotions, a comparative approach is necessary with our closest relative ones: the NHP (see Figure 12).

To simplify, we will divide the NHP in two main groups of species: the great apes (gorillas

whom we had a common ancestor around 14 million years ago and the monkeys, such as macaques or baboons, separated from the Hominidae branch 28 – 44 million years ago, depending on the species (Schrago & Voloch, 2013).

Figure 12: Primate phylogenetic tree with divergence time (in million years ago; Mya) based on comparative MRI and genotyping data. Numbers in parenthesis refer to the number of MRIs used for each species (Heuer et al., 2019).

Hence, several studies highlight the abilities of great apes and monkeys to express and perceive affective cues through faces, whole-body or vocalizations (Fragaszy & Simpson, 2011). For instance, using a matching-to-sample design, Kano and collaborators demonstrated the capacity of young female chimpanzees to recognize pictures depicting aggressive conspecifics more easily than pictures representing relaxed chimpanzees, considered as neutral (Kano, 2008). Interestingly, the authors also showed in juvenile and adult chimpanzees an attentional bias towards agonistic scenes compared to neutral and positive ones in naturalistic movies, involving the whole-body expressions of conspecifics (Kano & Tomonaga, 2010). This attentional sensitivity for negative affects has often been

demonstrated in the human literature (e.g. Kret et al., 2013; Sander, Grandjean, Pourtois, et al., 2005; Schaerlaeken & Grandjean, 2018), suggesting common mechanisms across primate species. Indeed, attentional bias towards threatening or fearful cues would maximize the chance of survival of an individual when reacting rapidly to a potentially dangerous situation. However, this attentional process would not be present in all primate species. For instance, in macaques (Macaca mulatta), Bliss-Moreau and collaborators showed an attention towards affective scenes involving whole-body conspecifics but they did not find any difference between aggressive and affiliative movies (Bliss-Moreau et al., 2017).

Furthermore, using a dot-probe task, a study in bonobos revealed an attentional bias for positive scenes involving conspecifics (Kret et al., 2016). These results could be explained by the nature of this peculiar great ape species. Bonobos are indeed a more social species than chimpanzees (Gruber & Clay, 2016) and thus, protective and affiliative behaviours could represent a pivot in their society, increasing bonobo attentional capture of positive affects.

Nevertheless, despite their more aggressive behaviour, chimpanzees are also capable of expressing affiliative contents through facial expressions identically to humans (Parr et al., 2005; Parr & Waller, 2006; see Figure 13). In addition, chimpanzees as well as bonobos, gorillas, orangutans and even siamangs (Symphalangus syndactylus), another ape species, are also able to laugh in playful situations similarly to human infants or adults (Davila Ross et al., 2009; Davila Ross et al., 2011).

a) b)

Figure 13: a) Prototypical facial expressions of laugh (left image) and smile (right) in human (adapted from Parr & Waller, 2006). b) The corresponding facial expressions in a young chimpanzee with play face (left image) and silent bared-teeth display (right) (Parr et al., 2005).

Finally, before closing this section on affects in animals, we will think further with this open question: Do animals understand and feel what others feel? In other terms, do they have empathy?

This question is currently debated among the scientific community. Yet, from an

the ability to express empathy through altruistic behaviours could be necessary to many species. In fact, literature on empathy in great apes often describes their probable ability to have empathy towards their conspecifics. For instance, researchers revealed the capacity of chimpanzees to console the victim of an aggression (de Waal, 2009). Similarly, they also demonstrated affective contagion in orangutans relying on facial mimicries (Davila Ross et al., 2008). Both of these behaviours seem indeed to involve the understanding of others’

feelings.

Despites controversial findings, some authors demonstrated possible empathic behaviours in more distant species such as birds or other mammals. For example, Fraser and Bugnyar suggested that ravens (Corvus corax), like chimpanzees, could show consolation behaviours towards the victim of an aggression, but only if the victim had a valuable relationship with the consoler (Fraser & Bugnyar, 2010). Moreover, some researchers also described possible compassion mechanisms in elephants (Loxodonta africana), detailing the peculiar interest of these elephants for the dead bodies of their conspecifics (Douglas-Hamilton et al., 2006).

The question of animals’ empathy is still open. Nevertheless, if we consider empathy as a simple perception-action mechanism that provides an observer (the subject) with access to subjective state of another (the object) through the subject’s own neural and bodily representations (de Waal, 2009; Preston & de Waal, 2002), we may hypothesize that empathic processing could be present in a large number of species.

Overall, affective scientists of the 21^st century, have by now accepted the usefulness of evolutionary thought for the general understanding of human emotions. Consequently, comparisons between emotional processes in human and affective processing in NHP are now often used. From this, section 2.2 will describe the evolutionary continuity between human (Section 2.2.1) and NHP’ affective vocalizations (Section 2.2.2).

2.2. Evolutionary continuity between Human and Primate Affective Processing 2.2.1. Emotional Prosody Recognition

Abilities to vocally express and recognize emotional cues in others are crucially involved in the survival of all primate species, humans included. Shaped by millions of years of evolution in its own lineage, the human species became an expert for the vocal expression and the recognition of emotions through prosodic modulations. Relying on specific

physiological changes and particular brain networks, the variation in acoustic features is indeed essential to emotional communication in human. Yet, while researchers mainly focused on conspecific voices (human to human), little is known about the human capacity to decode or not emotional cues in other species vocalizations. Fortunately, comparative studies on this matter recently emerged, especially on the human recognition of affect in non-human primate vocalizations. Interestingly, influenced by the familiarity or the acoustic proximity of the calls, humans seem able to identify affective valence or arousal in non-human primate vocalizations.

2.2.1.1. Emotional Prosody in Human Voice

Prosody refers to all suprasegmental changes in spoken utterance, including intonation, amplitude, envelope, tempo, rhythm and voice quality. Prosody through the modulation of acoustic features such as fundamental frequency (F0 – lowest frequency of the voice, pitch) or energy distribution, enables a listener to infer for instance, the identity or the emotional state of a speaker (e.g. Grandjean et al., 2006; Wagner & Watson, 2010). Prosodic modulation in the human speech is thus necessary to the encoding and decoding of emotional expressions.

In fact, studies demonstrate the link between acoustic parameter changes and the emotional valence or arousal expressed by humans. For instance, Elodie Briefer described that the arousal level of emotions correlates with modulations in the respiration and phonation, influencing acoustic features such as F0, duration or speech rate (number of syllables produced per time unit, velocity of speech). On the contrary, emotional valence relies on intonation and voice quality, mostly changing the energy distribution in the spectrum.

Positive emotions indeed show a lower energy in frequency compared to negative ones (Briefer, 2012).

Furthermore, specific prosodic modulations may refer to a particular emotion. For example, the literature often describes that i) angry voices rely on the increase of F0 mean, mean energy, high frequency energy and downward directed F0 contours (sequence of F0 values across the utterance including mean, start, end, maximum and minimum of F0, intonation contour); ii) fearful voices also correlate with an increase of mean F0, F0 range (difference between minimum and maximum of F0), high frequency energy, speech rate, jitter (cycle to cycle frequency variation of F0, pitch perturbation) and shimmer (cycle to cycle amplitude –

range, mean energy and downward directed F0 contours; and finally iv) happy voices also increase mean F0, F0 range, mean energy, high frequency energy, speech rate, jitter and shimmer (Banse & Scherer, 1996; Eyben et al., 2016; Goudbeek & Scherer, 2010;

Hammerschmidt & Jürgens, 2007; Juslin & Laukka, 2003). Overall, these findings emphasize the acoustic differences between high and low arousals in which higher arousal (anger, fear, happiness) correspond to an increase of the speech parameters and lower arousal (sadness) is linked to a decrease of these acoustic properties. Interestingly, the same authors also show that the way specific emotions are expressed acoustically is similar across cultures, suggesting inherited mechanisms (Zimmermann et al., 2013).

Evolutionary functions are indeed at play in the expression of acoustic features. For instance, current findings show that rough temporal modulations found in alarm screams were selectively used to communicate danger across signal types. Thus, the roughness of screams, in inducing intense fear, would confer a behavioural advantage in increasing the rapidity and the accuracy of a listener to decode emotional cues in screams (Arnal et al., 2015).

Physiological fluctuations in emotional states have been found to influence acoustic properties of the speaker’s voice in a reliable and predictable manner that is perceptually available to a receiver (Juslin & Laukka, 2003; Taylor & Reby, 2010). Well conceptualized in the source – filter theory of vocal production (Fant, 1960; Titze, 1994), this biological process is crucially involved in affective communication in humans as well as in other mammals (e.g. Briefer, 2012; Filippi, 2016; Fitch, 2000). The source – filter theory distinguishes two independent functions from different parts of the vocal apparatus: i) the source, including larynx, sub-laryngeal and laryngeal structures; and ii) the filter (or vocal tract) involving the tube connecting the larynx to the mouth and the nose from which a sound is propagated into the environment (Titze, 1994; see Figure 14).

Hence, studies have shown that the source with the rate of opening and closing the glottis as well as the length and mass of the vocal folds enable the F0 structure in the speaker’

voice. In addition, through muscular interactions and changes in airflow or sub-glottal pressure, the source is also able to shape the tempo, the duration and the amplitude of the vocalizations (Belin, 2006; Taylor & Reby, 2010). Thus, in modulating F0 or the speech rate for instance, the source plays an important role in the vocal expression of emotional valence and arousal in the speaker’s voice (Goudbeek & Scherer, 2010; Morton, 1977). Similarly, the

vocal tract (filter) length and the mouth position enable the production of multiple formants which are defined as the maximum of energy around specific frequencies (Belin, 2006;

Fitch, 2000; Taylor & Reby, 2010). For example, it has been demonstrated that in a positive context, humans often retract their lips while in agonistic situations they protrude their lips, modulating the speaker’ voice quality (Drahota et al., 2008).

Figure 14: Illustration of the source – filter theory of vocal production in goat, valid for all living mammals (humans included). The source sound is produced by the larynx by vibration of the vocal folds, determining the formants (F1 to F4), which correspond to a concentration of acoustic energy around particular frequencies (Briefer, 2012; Fitch, 2000).

Emotional vocalizations cover a broad range from complex speech utterance with elaborate prosodic features (Goudbeek & Scherer, 2010) to laughter (Davila Ross et al., 2009) or non-verbal affect bursts such as “ah” or “eww” (Scherer, 2009a), all relying on specific brain structures.

Belin and collaborators, demonstrated the specific involvement of the bilateral superior temporal sulcus (STS) in the perception of human voices. Including bilateral regions close to the temporal pole as well as anterior and posterior portions of the Heschl’s gyrus, these voice-selective areas showed greater neural activity for human speech and non-speech in comparison to environmental sounds (Belin et al., 2000).

Moreover, playing a key role in human voice perception, the STS and the superior temporal gyrus (STG) are crucially involved in the processing of vocal emotions. In fact, Grandjean and collaborators found a stronger activation in the middle STS (mSTS) and STG during the listening of pseudo-words expressed with angry prosody compared to the neutral ones. In a

Moreover, playing a key role in human voice perception, the STS and the superior temporal gyrus (STG) are crucially involved in the processing of vocal emotions. In fact, Grandjean and collaborators found a stronger activation in the middle STS (mSTS) and STG during the listening of pseudo-words expressed with angry prosody compared to the neutral ones. In a