Comparative thanatology, an integrative approach: exploring sensory/cognitive aspects of death recognition in vertebrates and invertebrates

Abstract

Evolutionary thanatology benefits from broad taxonomic comparisons of non-human animals' responses to death. Furthermore, exploring the sensory and cognitive bases of these responses promises to allow classification of the underlying mechanisms on a spectrum from phylogenetically ancient to more derived traits. We draw on studies of perception and cognition in invertebrate and vertebrate taxa (with a focus on arthropods, corvids, proboscids, cetaceans and primates) to explore the cues that these animals use to detect life and death in others, and discuss proximate and ultimate drivers behind their capacities to do so. Parallels in thanatological behaviour exhibited by the last four taxa suggest similar sensory–cognitive processing rules for dealing with corpses, the evolution of which may have been driven by complex social environments. Uniting these responses is a phenomenon we term ‘animacy detection malfunction’, whereupon the corpse, having both animate and inanimate attributes, creates states of fear/curiosity manifested as approach/avoidance behaviours in observers. We suggest that integrating diverse lines of evidence (including the ‘uncanny valley’ effect originating from the field of robotics) provides a promising way to advance the field, and conclude by proposing avenues for future research.

This article is part of the theme issue ‘Evolutionary thanatology: impacts of the dead on the living in humans and other animals’.

1. Introduction

It is often difficult to judge whether animals have any feeling towards others' sufferings. Who can say what cows feel, when they surround and stare intently on a dying or dead companion?

—Darwin [1, p. 73]

[Researchers] state without qualification that man is the only animal that can be aware of his own future death. But I suggest that we pause and ask just how anyone knows this. What sort of evidence is available either pro or con?. (…) The available negative evidence supports at most an agnostic position.

—Griffin [2, pp. 104–105]

Awareness of death has been regarded as one of the defining traits of Homo sapiens by distinct schools of thought, including philosophy [3], psychology [4], anthropology [5] and biology [6]. However, at least since Charles Darwin, the possibility of a psychological continuity between humans and other animals has been entertained. While Darwin himself expressed doubts regarding the extent to which non-human animals (hereafter: ‘animals’) could comprehend the death of a conspecific, his question was of a different nature from that of Griffin (see quotes above): he asked not if animals have an awareness of their own mortality, but, more modestly, to what extent they recognize death in others and how they ‘felt’ in response.

These questions are at the centre of the field of comparative thanatology [7], which is concerned with how animals respond to their dead. Recent scientific interest notwithstanding, so-called funerary activities among animals have been reported since ancient times. Most notably, stories of elephants and ants burying their dead or dolphins assisting dead companions to the surface are recounted by Pliny the Elder (AD 29–79) and Aelian (AD 175–235) [8,9]. Already in the eighteenth and nineteenth centuries, first-hand reports and anecdotes accumulated on the interactions of animals with dead conspecifics. These included protection of the corpse, transport, vigils and emotional distress, with allusions to grief in non-human primates [10,11], corvids [12,13], proboscids [14,15], cetaceans [16], ungulates [17], carnivores [18,19] and sirenians [20]. Observations similar to these have been confirmed more recently by researchers studying many of the same species [21–27].

In the present paper, our aim is to bring together hitherto partially disparate lines of research, which, in combination, can provide new perspectives on how, and to what extent, non-human animals detect death in others. We draw on empirical case reports as well as current research on animal perception and cognition relevant to elucidating some of their enigmatic responses towards their dead, compare these responses across taxa and discuss how they may have emerged in the course of evolution. We begin by exploring how animals detect life, then through what cues they might detect death, and which aspects of cognition might contribute to processing such sensory information as ‘death awareness’. We conclude by proposing avenues for future research.

2. Detecting life

Within their natural surroundings, animals are constantly confronted with dynamic (moving) visual signals. Interpreting these correctly is a significant factor in evolutionary fitness, as such signals can come from both living (self-propelled motion patterns) and non-living (objects moved by external forces) entities [28]. The ability to distinguish biological from non-biological movement was presumably part of an ancient mechanism that evolved for modulating interactions with other organisms, be they heterospecific predators or prey, or conspecific kin, mates or competitors.

Movement through self-propelled motion was present in the first living organisms during the Pre-Cambrian some 3000 Mya. These organisms moved around with the help of cilia and flagella [29]. Predation was already a strong selective force and one that likely gave rise to the first eukaryotes [30] around 1600–2500 Mya [31]. By the Ediacaran Period (635–540 Mya), a division between chordates, molluscs and arthropods appeared, as indicated by genetic and fossil evidence [32]. These organisms were essentially relegated to a two-dimensional world confined by bio-mat grazing [33] where, possessing photo-receptors, they navigated discerning between light and darkness. Later yet, during the Cambrian Period, in an increasingly three-dimensional world [34], innovations in many animals emerged that have continued until now: the presence of brains and nervous systems [35], fully formed eyes [36], attention [37], associative learning [38], embodied cognition [39] and even the beginnings of primary consciousness [40].

(a). Biological motion

It was in the field of biological motion perception [41] that researchers found evidence for a perceptual life detector likely common to all land vertebrates [42]. Humans [43,44] and newly hatched chicks [45] were presented with stimuli consisting of moving light point displays against a dark background, one set of stimuli depicting a moving being (human or other) and the other a moving but scrambled or inverted version of the first. Despite light points being severely impoverished percepts, the visual system decodes them in a straightforward manner when their movement corresponds to biological motion, but not non-biological motion; both studies found quick detection by humans and an inherent attraction in the chicks towards biological as opposed to non-biological motion. Further evidence for an ancient neural mechanism common for the detection of animacy came from additional studies with chicks [46] and with human newborns, who also attend preferentially to biological motion stimuli [47,48].

Recent work confirms the biological motion effect in adults of species such as Medaka fish [49], pigeons [50], rats [51], cats [52], dolphins [53] and non-human primates including common marmosets [54], rhesus monkeys [55], baboons [56] and chimpanzees [57]. But to date, no newborn non-human primate has been tested; hence, the issue of innateness is unresolved outside humans. The ability to detect biological movement paired with an inversion effect has some commonalities with other fields in perception such as face recognition in humans [58] and non-human primates [59], a capacity that appears to be innate in many species including primates [60]. The perception of biological motion is likely only one of many components of a larger perceptual system for animacy detection comprising ‘detectors’ of biological cues such as faces, eyes, texture, odour and particular shapes, all fundamentally tied with agency attribution. In humans, self-propelled motion by itself is not a sufficient cue to trigger detection of intentional agency [61,62].

(b). The animate/inanimate distinction

In a recent review of the development of the animate–inanimate distinction in human infants, in addition to biological motion and self-propelled movement, Opfer & Gelman [63] list goal-directed movement (the directness with which an agent moves towards its perceived goal) and contingency of behaviour (the timing between an agent's actions and specific events) as cues. In studies of human infants' capacity to decode an agent's intentions, infants seeing a human hand reaching for an object react to changes in its goals, whereas no such response is observed when a mechanical rod or claw replaces the hand [64–68]. Similar results have been found in capuchin monkeys (Cebus apella) [69], while the use of a monkey-like robot but not a moving box induces goal-direction ascription in common marmosets (Callithrix jacchus) [70]. During later development, these components of animacy attribution underlie various aspects of human infants' social cognition [71]. Importantly, the developing refinement of understanding and attribution errors are informative in terms of infants’ categorization of ‘alive’ versus ‘dead’ (including discrete components of full death-awareness: universality, irreversibility, causation and non-functionality [72]; see also Anderson [73] this issue).

The animate and inanimate conceptual categories may relate to distinct neural circuits representing domain-specific knowledge systems that are evolutionarily adaptive [74,75]. Support for this theory includes discovery of a close match between humans and rhesus macaques (Macaca mulatta) in inferior temporal cortical object representations, both categorical and continuous [76,77].

3. Detecting death

(a). Scent cues

In the animal kingdom, responses to dead conspecifics include necrophobia (avoidance), necrophagia (feeding) and necrophoresis (transport), and in many cases, these are elicited primarily by chemical signals. Aversion to ‘death scents’ (fatty acid necronomes or cadaverine/putrescine) may be a highly conserved response that is either innate or acquired easily due to a predisposition. Unlike freezing [78] or thanatosis (death-feigning) [79], which are visual in character, energetically non-costly and have evolved to transmit a specific message from sender to receiver, scent associated with decomposition is an unambiguous cue that receivers can reliably exploit. Necrophobic responses are adaptive in terms of predator evasion or pathogen avoidance. Below we explore in more detail specifically scent-triggered responses to dead conspecifics across different taxa.

In eusocial insects (numerous species of ants, termites and bees), two major, non-mutually exclusive scent-based hypotheses have been proposed to explain the typical burial or corpse removal responses to dead colony members (reviewed in [25] and [41]; see also Sun et al. [80] this issue). These hypotheses relate to the presence of specific chemical death-cues (necromones) and to the absence of chemical signatures of life (chemical vital signs), respectively. For example, while some ant species apparently respond to the decomposition-driven accumulation of fatty acids in corpses [81,82], others, such as the Argentine ant (Linepithema humile), engage in undertaking activities following the rapid disappearance upon death of cuticular chemicals secreted by live individuals [83]. As the latter mechanism potentially allows for a faster response, it is likely favoured in situations where the removal of dead bodies is time-sensitive, for example, in densely populated colonies where the risk of contamination from pathogens is high. Interestingly, sensitivity to these cues appears to vary in some species according to caste—e.g. soldiers of Atta mexicana do not respond to oleic acid, the most common death cue in many social insects [84]—consistent with the observation that undertaking and corpse removal responses are often performed by a subset of colony members only, in a division of labour. Outside of eusocial insects, several genera of Isopoda as well as social caterpillars are known to be sensitive to oleic and linoleic acid extracts (and to avoid these chemical cues), suggesting an ancient origin stretching back to at least the Crustacea–Hexapoda common ancestor 420 Ma for the involvement of necromones in arthropod corpse removal behaviour [85].

By contrast, among vertebrates, death cues such as cadaverine and putrescine, alongside other decay- and putrefaction-related substances, typically elicit aversion responses. Zebrafish avoid cadaverine [86] and show elevated stress levels upon encounter [87], and sea lampreys [88] and sharks [89] avoid odours emanating from decaying conspecific tissue. Rodents, including various species of mice, voles, shrews and chipmunks, also avoid areas where deceased rodents—even heterospecifics—are present, presumably at least partly informed by olfactory cues associated with decaying flesh [90]. California sea lions reportedly avoided a pool used for cooling after the carcass of a dead pup fell into it, moving away after apparently sniffing at it [91]. In humans, a range of interesting emotional and conscious, and unconscious responses to putrescine have been documented, including increased vigilance, active and planned retreat, and hostility towards out-group members [92]. However, not all vertebrates show avoidance: in rats, cadaverine and putrescine elicit the burying of dead conspecifics [93], and in goldfish, the same chemicals enhance feeding activity [94].

(b). Beyond scent: visual, tactile, multi-modal cues

Many animal species exhibit complex responses towards their dead that are not necessarily triggered by scent; furthermore, these often rely on a combination of several sensory modalities. Already among the arthropod examples described above, scent cues may combine with tactile cues to modulate responses to dead conspecifics (e.g. in the termite Reticulitermes virginicus; [95]), although the latter alone are insufficient to trigger a response. Here, we explore non-olfactory and multi-modal cue use in corvids, proboscids, cetaceans and primates, with a focus on observational and experimental field studies. The species comprising these taxa display many complex behaviours across both the physical and social domains, and have been argued to possess episodic-like memory. Many live in hierarchy-based social structures in which they cross-modally recognize individuals and act based on their past interactions [96–102]. Moreover, some individuals in these taxa are able to recognize themselves in the mirror, suggesting self-awareness [103–106]. It has been suggested that such species might be capable of an understanding of death [107]; however, there appears to be no qualitative difference between species that fail and those that pass the mirror test of self-recognition in regard to thanatological behaviour (i.e. dead infant carrying, exploratory behaviours towards the corpse, vigils, visitations, etc.) [108].

(i). Corvids

The family Corvidae includes crows, ravens, rooks, magpies, jays and jackdaws. These species generally live in bonded pairs, possess the largest relative brain size of any avian group and show rates of behavioural innovation and complexity unparalleled in other bird species [109].

Observations of thanatological behaviours have been made in several corvid species. Several reports describe a ‘ceremonial gathering’: an assembly of living individuals near a deceased conspecific. The participants utter alarm calls but seldom touch the corpse or show aggression, in comparison to their predator mobbing or scavenging the corpse of another species [110–114]. Also, compared to the amount of time spent by cetaceans, elephants and primates near corpses (see below), these gatherings tend to be relatively short-lived. There are two reports of objects (feathers, sticks/grass) being placed near the corpse [111,112]. Survivors subsequently tend to avoid the place where a dead conspecific is found [115], so much so that effigies have been found useful for pest control [116,117].

In many bird species, a dead conspecific generates cautious inspection and/or mobbing behaviours (reviewed in [118]). Lorenz [119] reported his tame jackdaws (Corvus monedula) attacking him when he carried his black swimming shorts in his hand, and suggested that the likeness of the trunks to a dead jackdaw triggered this mobbing behaviour. Barash [120] paired a predator model—a great horned owl (Bubo virginiatus)—with either a black cloth or a crow model and obtained similar results: live crows (Corvus brachyrhynchus) mobbed these significantly more than the owl model alone. Feathers resembling those of a conspecific may also trigger alarm responses in several crow species [121–123].

The primary mode of recognition, therefore, is likely visual: corpses not exhibiting visual cues such as coloured feathers will not elicit responses from live conspecifics. For instance, Heinrich [110] described how a dead crow he attempted to feed to live ravens (Corvus corax) was promptly rejected; it was only accepted as food after removal of the feathers, head, wings and feet. Similarly, Lorenz [119] found that adult jackdaws did not mob him if he was holding a young jackdaw before it acquired black feathers, but they did after those feathers grew. In what has been termed the information-gathering hypothesis [7], assemblages around a dead conspecific might serve to acquire information surrounding the death and to assess a potentially dangerous situation. Additionally, assembling corvids might be appraising new social changes in the group [111,113,115,118,124]. Three experiments formally tested these hypotheses [115,124,125]. In scrub-jays (Aphelocoma californica), corpses of conspecifics and similar-sized heterospecifics elicit aggregations and site avoidance [125]. A corpse in prone posture elicits cacophonous aggregations, whereas an upright one elicits mobbing behaviour [124] (see also Swift & Marzluff [126] this issue). Moreover, unlike dead conspecifics, corpses of pigeons (Columba livia) elicit little reaction in crows. Interestingly, pigeons are similarly low-responsive to dead pigeons, suggesting that this species processes and evaluates the situation differently from crows [115]. Earlier research on wood pigeons (Columba palumbus) used models and showed that pigeons tend to avoid corpses of conspecifics as a default response [127,128], which contrasts with corvids' initial curiosity.

Studies on crows have not yet explored how the social relationship with the dead affects the interactions of the living, particularly in the case of pair-bonding corvids, although this has been alluded to in single-case reports [111,114] (see also §5). In addition, compared to the taxa discussed next, corvids show limited prosocial tendencies [129], including little regard for conspecifics beyond kin or mate; this may explain their shorter-duration and more frequently agonistic responses to corpses compared to non-human primates, proboscids and cetaceans.

(ii). Non-human primates

Currently, more is known about thanatological behaviours in non-human primates than any other vertebrate taxon, and these include mobbing/alarm calling, aggression, dead infant carrying, vigils and visitations of the corpse. Physical interactions with the corpse include grooming, gentle touching, poking, attempted sexual mounting, dragging, rough touching, hair-plucking, beating and even cannibalism (reviewed in [108]; see also Anderson [73] and Watson & Matsuzawa [130] this issue).

Dead infant carrying is a prevalent behaviour throughout Old and New World anthropoid primates, and can last from a few hours to a few months. Reports of extended carrying behaviour exist for great apes [131–134] and Old World monkeys [135–138], with the largest study, on Japanese macaques, documenting 157 cases over a 24-year period [139]. Several hypotheses have been proposed to explain such behaviour, not necessarily mutually exclusive: infants live or dead are perceptually attractive to females: maternal hormones involved in mother–infant bonding likely reinforce carrying behaviour, and some behaviours (such as removal of larvae by grooming) and climatic factors (such as high altitude or low humidity) can contribute to preserving the corpse for extended carrying (reviewed in [108]). Contrary to the mother, other adult group members show little interest in the dead infant, engaging much more with live ones [136,140–143], although periodic inspections and attempts to play with and carry a mummified corpse (e.g. [144]) have been reported. Instances of guarding the infant corpse or its mother against approaches by other group members have also been witnessed [143,145–148].

To what extent primates’ (or indeed any taxon's) responses to the dead are shaped by learning is an intriguing question. Witnessing death events can allow information-gathering about various sets of cues associated with the phenomenon that can be retrieved on similar subsequent encounters; active information seeking in such situations has been suggested, for example, in chimpanzees [149]. With respect to the extended carrying of dead infants, the eventual abandonment of the corpse by a mother may reflect her acquisition of some component of death-awareness, through a combination of visual, olfactory and behavioural cues—or their lack of correspondence with those emitted by live infants. Social learning may also shape some thanatological responses: it has been suggested that witnessing dead-infant carrying by others may promote the behaviour in mothers experiencing their own infant's death [144].

Experimental studies of thanatological responses in wild primates are largely lacking in comparison with corvids. Allen & Hauser [150] proposed an experiment using playback calls (as done with vervet monkeys (Chlorocebus aethiops)) with recently dead infants, to study concept attribution (e.g. cognizance of death) in non-human primates. Females, they argued, when presented with the playback would (i) orient towards the speaker and act as if the infant were alive, (ii) respond in a distressed manner and search for the infant, or (iii) not react at all and continue engaging in ongoing activities. To our knowledge, this experiment has never been implemented, possibly due to ethical concerns. However, a variation was conducted in a study of the strength of male–female relationships in free-ranging chacma baboons (Papio ursinus) [151]. The authors found that males responded to a female's call if there had been a close association (friendship) between them, but not if the female's infant had recently died. Their interpretation was that females, primarily responsible for maintaining these close associations, ended friendships with males upon their infant's death, either because they no longer needed a male to protect their infants or because they themselves no longer benefited from protection by the male. However, an alternative interpretation is that the males themselves chose not to respond on the basis of knowing that the infant had died (including, perhaps, knowledge of the event's irreversibility). We further discuss potential experimental approaches for studying death-related psychological states in non-humans in the final section.

(iii). Proboscids

The order Proboscidea comprises three extant species, the Asian elephant and the African bush and forest elephants. Wild elephants live in complex fission–fusion societies with female matrilineal kin forming a family unit with close, lifetime bonds. The encephalization quotient of elephants rivals that of primates, and they possess as many cortical neurons as humans do, albeit less densely packed than in primate brains (reviewed in [99,152]).

Like non-human primates, elephants have been observed to surround a dead conspecific, interact directly with it, touch it with their feet or trunks, at times attempt to lift it with either foot or tusks, and vocalize in apparent distress. They may also guard the body against predators or other conspecifics and revisit the corpse in the following days [153–156]. Adult females have also been observed carrying dead infants weeks after death [157–159].

Unlike non-human primates observed to date (but see [160]), elephants occasionally cover dead conspecifics with branches, leaves or soil, and may attempt to patch wounds on the dead with dirt or put food in their mouth [21,22,153,158,161]. It is important to consider these behaviours in the context of the elephant's social repertoire with live conspecifics and heterospecifics. Elephants have reportedly buried humans and dead animals [162,163].

Contact behaviours with the corpse include using the trunk to inspect the head and body; even the torso may be used for such inspections. Pulling and stepping over the corpse have been observed, as have scent-related behaviours such as sniffing the corpse with the tip of the trunk and displaying the flehmen response (touching the tip of the trunk to the openings of the vomeronasal organ). Elephants may also put the trunk in their mouths to assess gustatory information about the corpse [156,164]. Elephants show striking responses to the bones of other elephants, particularly skulls and tusks, carefully inspecting them [21,22,165]. McComb et al. [166] showed experimentally that African elephants are primarily attracted to tusks in comparison to skulls, pay more attention to conspecific skulls than other objects and show no evidence of recognizing skulls of familiar conspecifics. The attraction to tusks might be because they represent an externally visible cue to identity that is consistent across life and death.

Playback of calls of dead elephants to live group members has also been attempted [167]. When vocalizations of a female were broadcast to her family unit 3 and 23 months after she had died, group members responded with contact calls each time, even approaching the speaker (but did not do so in control trials involving the vocalizations of unfamiliar individuals), suggesting long-term memory and recognition. As it was not specified whether the group members had directly observed the death or seen the female's corpse, it is unknown whether responses to playback calls would differ depending on such knowledge.

(iv). Cetaceans

Thanatological behaviour among cetaceans (whales, dolphins and porpoises) is also becoming increasingly well documented and shows many parallels with primate and proboscid data. The vast majority of reports (compiled in Reggente et al. [168], see also [27] for a recent review) concern interactions with dead calves or juveniles; carrying their carcasses has been documented in various dolphin and whale species. Indeed, the behaviour has been observed worldwide and in a range of environments including open oceans, bays and inlets, and rivers [169]. Although carrying can be for extended periods, due to the nature of the aquatic environment rapid decomposition limits carrying duration in comparison to, for example, primates in dry habitats. Carrying typically involves holding the calf on the dorsal fin, against the melon or in the mouth. Along with transport, potentially breathing-related behaviours such as lifting the corpse to the surface of the water and pushing it down have been observed (e.g. [170,171]).

Aside from transport, several other categories of behaviours have been documented, including striking the corpse, non-contact attendance such as stationing next to the corpse, and sexual arousal and copulation (towards adults only; e.g. [172]). Unrelated individuals also occasionally interact with corpses, and carriers of an infant corpse are frequently seen surrounded by other pod members swimming in cohesive, possibly protective formations [169].

Like proboscids, cetaceans possess a keen sense of hearing that likely plays an important role in navigating their physical and social environments [173]; however, most cetaceans do not possess a sense of smell or taste [174]. This is likely to impact both the sensory drivers and the nature of their interactions with the dead. Visual cues (presence of wounds, lack of autonomous movement) and lack of auditory cues (vocalizations) are the most probable sources of information about a deceased conspecific's state.

(v). Vertebrate species in context

Outside of the taxa discussed above, various mammalian species including giraffes [175–178], otters [179], dingoes [180], seals and sea lions [114,181–183] and manatees [184] have been observed stationing around, manipulating, or carrying their dead infants for extended periods of time. Phylogenetically ancient maternal caretaking mechanisms continue to operate even after the offspring has died in both mammalian and avian taxa. On a proximate level, the mother may perceive the infant's condition as ambiguous, or she may anticipate that the infant will yet recover, whereupon she continues her caregiving. On an ultimate level, her actions likely represent behavioural error because of the cost of too readily abandoning a potentially temporarily unresponsive infant. A transitional phase ensues that can vary widely (days, weeks or months), during which the mother will carry or stay in close proximity and interact with the corpse (e.g. inspecting, grooming, licking); these responses will decrease over time, culminating in abandonment or occasional consumption of the corpse (see below).

Nonetheless, corvids, primates, proboscids and cetaceans appear to exhibit the greatest similarities in thanatological behaviours, a trend we predict to occur in other behaviourally/socially complex taxa (see e.g. [185]). This is surprising as they do not share a recent evolutionary past and occupy different ecological niches. What they do have in common, however, are complex societies, extended parental care and large brains. Hence, the parallels among these taxa in thanatological responses may be the result of similar perceptual–cognitive processing rules that evolved in the context of increasingly complex social environments. Responses to adult conspecific corpses are both stronger and longer than in other taxa, in which the most common response is avoidance (e.g. rodents [90]). Alternatively, the fewer occurrences of thanatological behaviours reported in other vertebrate species could be due to observation bias (see §5 for further discussion).

4. The ‘uncanny’ corpse

What are the perceptual–cognitive processing rules that give rise to complex thanatological responses? Here, we draw on several aspects of visually oriented animals' detection of dead conspecifics to propose a novel synthesis of underlying cognitive mechanisms. Species with larger brains and more advanced cognitive processing, causal reasoning, and information-gathering abilities appear to have comparable responses, suggesting an overlapping phenomenon that is shared across them.

(a). Threat assessment mechanisms

Brains coupled to nervous systems evolved as a means to process ecologically relevant information, and to orchestrate adaptive interactions with the surrounding world. They emerged to deal with the challenges that arise from the physical and social environments, and as these became increasingly complex, so did organisms and their brains [40,186,187]. Detecting cues to the presence of life-threatening risks remains critical for animals, and natural selection has equipped organisms with and without large brains to do this. However, with associative learning, animals no longer adapted only through evolutionary time, but also within their lifetimes via experience-based behavioural adjustments. Thus, cognition and memory, capacities shaped by natural selection, are critical in regulating expectation, detecting discrepancies and anticipating events.

The corpse, a highly salient object, represents a conceptual novelty (any familiar object displayed in unfamiliar configurations or unfamiliar settings) (sensu [188]). Comparative neuroimaging research has revealed that novel stimuli are encoded by the hippocampus [189] and the amygdala [190], and damage to these areas diminishes fear and vigilance to threat (reviewed in [191]).

Cross et al. [192] used positron emission tomography scans to examine cerebral circuitry involved in integrating visual cues into behavioural responses in crows. Crows possess brain regions analogous to the hippocampus and amygdala in mammals that are activated during potentially dangerous situations. The sight of a novel person holding a dead crow activated visual pathways and the hippocampus, while the amygdala was significantly activated by a predator stimulus (a hawk). These patterns of activity were explained as distinct processing activities when gathering novel threat information (person holding a dead crow) versus retrieving past fear information (mounted hawk).

(b). The uncanny valley

Corpses, as passive entities, defy expectations, provoke emotions and generate various behaviours in the living. Notably, they present a conflicting mixture of presence (odour, wounds, vermin) and absence of cues (movement, sound, body heat). This contradiction is illustrated by the dual approach/avoidance and exploratory/fearful reactions when encountering a dead body. Paying attention towards dead conspecifics is, as previously discussed, evolutionarily relevant, because the corpse might provide information about potential predation events or a pathogen hazard. Furthermore, taking a proximate approach, the living may be responding to novelty in the form of something ‘uncanny’.

The uncanny valley phenomenon was originally proposed by roboticist Mori [193,194] to describe the eerie feeling humans experience upon encountering a human replica, and we suggest that it also applies to thanatological responses in non-human animals. Mori used examples such as hand replicas and dead bodies to describe the drop in emotional valence the closer something resembles living specimens of our own species, movement being a key factor in the intensification of eeriness. Cognitive hypotheses posit that an uncanny eliciting stimulus remains in a category boundary or provokes a perceptual mismatch, two explanations that are not necessarily mutually exclusive [195] and not necessarily related to cadavers. An interesting example of this is Goodall's description of chimpanzees' fearful and aggressive responses towards physically deformed conspecifics affected by poliomyelitis who moved in unusual ways [196]. At the ultimate level, an adaptive pathogen-avoidance mechanism could be at play, whereby abruptly acquired physical abnormalities in conspecifics trigger a disgust response in other group members (sensu [197,198]).

Steckenfinger & Ghazanfar [199] attempted to test the uncanny valley effect in rhesus macaques and found that both humans and macaques display the same aversion to realistic reconstructions of conspecific faces, particularly, as predicted by Mori, if these were moving (also see [144] for an example of aversion to a ‘moving’—i.e. dragged—corpse among wild chimpanzees). This suggests that the mechanism causing uncanny-valley-like responses was present already in the common ancestor of Old World monkeys and H. sapiens. Experimental work with human infants suggests that the uncanny valley phenomenon emerges in the first year of life, likely due to perceptual narrowing and learning/differentiation processes [200]. Regarding the quality of dynamic cues, research suggests that the more natural movement is (see §2a), the more likely it is to be accepted by human subjects (suggesting less of an uncanny effect) [201]. Some types of stimuli such as androids or corpses likely fail sensory/cognitive scrutiny based on these learning processes, thus triggering an aversive response.

(c). Animacy detection malfunction

In primates, life detection is part of a series of core knowledge systems [202], in this case, the core system of agency (C.S.A.). The core knowledge theory proposes that hard-wired cognitive skills shape mental representations about the world. It remains unclear how many subsystems contribute to agency representation and how they are inter-related; however, some have been unveiled by developmental and comparative cognitive scientists [202–204]. The animate/inanimate distinction level likely operates through dual core knowledge systems specialized for dealing with animate and inanimate entities: the aforementioned C.S.A. and the core system of object (C.S.O.), respectively. Contrary to agents, objects are predictable. They are inert, moving only when external force is applied to them, and as such, they exhibit no contingency or any of the other traits associated with agency: they neither act nor react, but are acted upon by the living agents.

We suggest ‘animacy detection malfunction’ as a cognitive term for the conflicting responses exhibited by vertebrate taxa upon encountering dead conspecifics. The agency system is not perfect but prone to error; however, in terms of its primary purpose (agent detection), it normally functions well. Inspired by earlier views [194,205,206], animacy detection malfunction is defined as the conflicting cognitive process upon seeing a corpse brought about by perceptual mismatch ultimately causing a violation of expectation. The mismatch stems from the absence of dynamic cues to animacy with the presence of static cues to animacy and is intensified by individual recognition of the dead conspecific. The corpse then has both animate and inanimate attributes, triggering a conflict between the core knowledge system of agency and the core knowledge system of object.

(d). Death detection mechanism

Humans have long dealt with conflicting stimuli from corpses through cultural mortuary practices that are rooted in the deep hominid past [207]. Barrett & Behne [208] argued for the existence of a death detection mechanism, evolved through the course of human evolution, contending that reliable visual cues indicating death, such as fatal disruptions of the body envelope (e.g. decapitation, severe mutilation), were important in the recategorization from ‘living’ to ‘dead’ in humans. This can be illustrated with reference to predator detection accuracy, where failure results in death. The ability to discriminate a live predator (snake, leopard, crocodile) from a dead one allows for the activation of different behavioural decision-making outcomes with implications for survival [208].

In a study of grief after the loss of a companion animal, White et al. [209] found that humans viewing a corpse that exhibited reliable cues for death (i.e. grievous injuries, disruption of the body envelope) displayed less vigilant behaviour than when the corpse was intact. What they termed ‘false recognitions’ (incorrect attributions of sight and sound to the deceased) were also more frequent when the corpse was intact. Adopting and expanding on Barrett & Behne's [208] death detection mechanism, these authors suggest that natural selection shaped the increase in vigilance behaviour whenever a valuable partner was missing, and that attending to reliable cues of death was selected for throughout human evolutionary history.

Earlier applications of these assumptions (i.e. uncanniness, bodily disruptions) feature in experiments on fear performed by Hebb. He revealed concealed objects in a box to captive chimpanzees. Among the stimuli were what he called ‘primate objects’, which included a plaster taken from the death mask of an adult female chimpanzee, adult and infant chimpanzee replicas, an adult human head replica, a juvenile chimpanzee skull with a movable jaw, the mounted skin of a spider monkey with movable head and shoulders, the curated hide of a juvenile chimpanzee and the corpse of an infant chimpanzee. Some of the objects elicited intense fear or panic (in decreasing order: movable chimpanzee skull, snake cast, movable spider monkey skin, chimpanzee death mask), which Hebb interpreted as fears due to conflict; he suggested that the sight of mutilated bodies aroused an incompatibility at both the perceptual and at the cognitive level [205].

Butler [210] tested rhesus macaques in a test-box where they could see through an opening into another box. This other box contained live snakes, a live conspecific, an anaesthetized conspecific or a dead conspecific with its head on its outstretched hands. Butler predicted that the more frightful the stimulus, the more suppressed the viewing behaviour would be. However, this was not the case, and Butler explained the monkeys' reactions as a possible result of a psychological barrier between the subjects and the objects because no physical contact was possible. The decapitated monkey did elicit more looks than the live one, even though mean looking times were higher for the latter. This result might reflect a configural incongruity in the corpse eliciting a greater number of viewings but decreased overall looking time due to aversion (sensu the uncanny valley phenomenon).

How do these considerations help us advance our understanding of death awareness in non-humans in their natural environment? Boesch [160] has suggested that wild chimpanzees have a capacity for the ‘causation’ subcomponent of a full-blown awareness of death. Chimpanzees of the Taï Forest (Ivory Coast) face higher predation risks than many other chimpanzee communities [147]; they exhibit more fearful responses to individuals that died of disease (10 cases) than those that show wounds due to leopard predation (5 cases). Furthermore, Taï chimpanzees lick the wounds of injured group members, but not the dead. If the reason for these differences lies in an understanding of reliable cues for death (grievous wounds, severe disruptions of the body envelope), then chimpanzees may have an implicit awareness of death, not only distinguishing between dead and alive, but also between different manners of death, potentially providing evidence for the subcomponent of causation.

5. Conclusion and future work

Our review has brought together observations of living individuals' responses to dead conspecifics in invertebrates and vertebrates, evidence regarding the sensory bases of detecting life and death in others, and potential cognitive underpinnings for animals’ awareness of death. We suggest that phylogenetically ancient responses relating to death that are present in many animals exist not only for specific predator detection but also form part of a generalized threat detection mechanism. Presumably in corvids, cetaceans, proboscids and non-human primates, these mechanisms run in parallel with living–dead discrimination processes based on associative concepts. We also argue that analogical reasoning is a sine qua non condition for human-like death awareness with all of its main subcomponents (universality, irreversibility, cessation and causation).

Many authors have called for more and better data on animals' responses to the dead (including a number of contributors to the present issue—e.g. Watson & Matsuzawa [130]; Reggente et al. [168]; Anderson [73]) to advance comparative thanatology—greater taxonomic breadth, more quantitative descriptions and more systematic phylogenetic comparisons. While fully supporting these calls, we also advocate controlled experiments to probe the sensory and cognitive bases of the detection of death and its associated psychological states. For example, presenting taxidermy specimens of dead individuals of various species and in various poses might elucidate what cues trigger responses to the dead, and what adaptive explanations might lie behind animals’ interest in the dead [115,124,125]. In addition, further manipulation of stimuli—such as the computer-generated images used by Steckenfinger & Ghazanfar [199] to probe the uncanny valley effect in monkeys—could permit analysis of underlying cognition. What cognitive processes are tapped into when detecting life and death in others? How are cues that conflict within or between modalities (e.g. a decapitated but moving individual, or a moving individual smelling of necromones) resolved in different taxa, and can cross-species comparisons of cue hierarchy inform our reconstruction of phylogenetically ancient versus derived mechanisms for death detection? How do parameters such as state of decomposition, visible cues indicating cause of death, social/kin relationship of corpse and observer, observers’ previous experiences with death, etc., influence responses in different taxa? Of course, such experiments need careful ethical consideration to minimize distress to subjects.

In addition, playback experiments like those proposed by Allen & Hauser [150] could probe how living individuals conceptualize dead conspecifics, stress assessment (such as analysis of glucocorticoid levels [211]) can reveal physiological reactions to the loss of conspecifics with close or distant social or kin bonds, and non-invasive neuroimaging studies [192] might demonstrate how animals process corpses on the animate–inanimate spectrum. While both technically and ethically challenging, such a research programme may go a long way towards elucidating the proximate and ultimate drivers of thanatological responses across taxa.

Data accessibility

This article has no additional data.

Authors' contributions

The authors conceived and wrote the manuscript together.

Competing interests

We declare we have no competing interests.

Funding

We received no funding for this study.