Why chimpanzees carry dead infants: an empirical assessment of existing hypotheses. Elizabeth V. Lonsdorf, Michael L. Wilson, Emily Boehm, Josephine Delaney-Soesman, Tessa Grebey, Carson Murray, Kaitlin Wellens and Anne E. Pusey. Royal Society Open Science, July 1 2020.
https://doi.org/10.1098/rsos.200931
Abstract: The study of non-human primate thanatology has expanded dramatically in recent years as scientists seek to understand the evolutionary roots of human death concepts and practices. However, observations of how conspecifics respond to dead individuals are rare and highly variable. Mothers of several species of primate have been reported to carry and continue to interact with dead infants. Such interactions have been proposed to be related to maternal condition, attachment, environmental conditions or reflect a lack of awareness that the infant has died. Here, we tested these hypotheses using a dataset of cases of infant corpse carrying by chimpanzees in Gombe National Park, Tanzania (n = 33), the largest dataset of such cases in chimpanzees. We found that mothers carried infant corpses at high rates, despite behavioural evidence that they recognize that death has occurred. Median duration of carriage was 1.83 days (interquartile range = 1.03–3.59). Using an information theoretic approach, we found no support for any of the leading hypotheses for duration of continued carriage. We interpret these data in the context of recent discussions regarding what non-human primates understand about death.
4. Discussion
In this contribution, we presented the largest systematic and quantitative study of variation in infant corpse carrying in wild chimpanzees. We found that chimpanzees carry infant corpses at high rates, despite behavioural evidence that they recognize that death has occurred. Moreover, none of the variables we examined to test existing hypotheses for the duration of ICC were significant.
Recent discussions surrounding the phenomenon of infant corpse carrying have recommended reporting the rate of corpse carrying [
18] and what factors contribute to variability in rates [
16]. We found that in every recorded case in which a mother was able to access her infant's body, she carried it. The only exceptions to this pattern were when the infant was killed by infanticide and the mother never regained access, and when an infant corpse was recovered in the field and the mother was not seen (with or without the corpse) in the interim. It is possible that the disappearances represented cases in which infants were not carried, but given that we cannot see every chimpanzee every day, it is also possible that they were all carried. Even if we assume that all of the 41 individuals that disappeared were not carried, and include the two cases in which corpses were recovered without ever being seen with their mother, the carrying rate would be 50/93 or 54%. What is clear is that rate of carrying we report here, between 54% and 100%, is substantially higher than that reported in Japanese macaques in which 15% of all corpses are carried and 28.7% of very young infants (less than or equal to 30 days old) are carried [
18]. However, complete information regarding what happened to macaque infants that were not carried (i.e. whether corpses were observed but abandoned immediately or whether the infant simply disappeared/was subject to predation, etc.) was not reported. Such information would allow us to more fully understand the differences in carrying rates we report here. Comparable rate data are not yet available for other primate populations, and should be a focus of ongoing data collection.
We documented 33 cases of dead infant carrying, ranging in length from 30 min to over 15 days. On average, infant corpses were carried for a little over 3 days, and only two were seen to be carried over 10 days. Both of these carriers resumed cycling and mating while still carrying the corpse. In 24% of instances, the primary carrier was a non-mother but there was no difference in the duration of carriage between mothers and non-mothers. Interestingly, one of the relatively prolonged cases was performed by a non-mother who abducted a live infant and then carried it after it died. The mean carrying duration we found for Gombe was shorter than previously reported for wild chimpanzees. A recent meta-analysis of cases in anthropoid primates [
17] reported a mean duration for chimpanzees of 32.94 days, as well as the largest within-species variance among the other species examined, from 2 to 114 days. These authors included only chimpanzee cases extracted from previously published reports (
n = 9), so it is possible that the published literature is biased towards extremely long cases. However, a compilation of 14 cases from Mahale [
27] also reports a larger range of durations, from 30 min to approximately 126 days. An interesting possibility is that there may be site-specific cultural differences in responses to death [
2,
21] as there are with several other behaviours [
51]. Alternatively, differences in ranging patterns may affect duration of carriage such that larger day ranges constrain duration of carriage, as suggested by Carter
et al. [
52] for chacma baboons (
Papio ursinus).
Several proximate explanations have been put forth to explain either the likelihood or duration of ICC. We examined infant age at death to test the post-parturient condition hypothesis and the maternal-bond strength hypothesis, but none of our candidate models that contained infant age at death provided a better fit to our data than the null model (
table 3). Moreover, we documented several instances of carrying by non-mothers as well as two cases of prolonged carrying in which the carriers resumed cycling and mating while still carrying the corpse (as also seen in geladas [
23]). The only notable age-related pattern we found was that no infants over 3 years of age were observed being carried. However, there were only nine individuals between the ages of 3 and 5 years of age in the full dataset, five of which disappeared (see electronic supplementary material, table S2). Of the remaining four, one was orphaned, one was killed by chimpanzees, one was killed by humans, and one died of illness. Thus, only the infant that died of illness could have been carried, but there were 6 days between sightings of this mother with and without the infant, so carrying cannot be determined. One possibility is that at 3 years of age, which is when most offspring cease riding on their mothers [
50], corpses are simply too heavy and/or awkward to be carried for lengthy periods of time. A similar age effect was reported in Japanese macaques, in which carrying rates increased with age up until 30 days, but then decreased [
18]. In sum, the small number of deaths above age three and the uncertainty around them prevent determination of whether there is an upper age limit for an infant to be carried, but the combined evidence does not support the post-parturient condition hypothesis of infant carrying. Bond strength probably plays a role in
who carries a corpse given that mothers, adoptive mothers and siblings exhibited the most carrying (
table 4), but it does not appear to predict duration of carriage in a consistent way.
The slow decomposition hypothesis [
23] posits that prolonged infant carrying is facilitated by dry climates and/or in the dry parts of the year. According to our model comparison, the model including season was the second-highest-weighted model (
table 3), although this model performed no better than the null model. In fact, while our two longest cases (
table 2) occurred during the dry season, carrying duration tended to be longer in the wet season (though this parameter estimate included 0 in the 95% CI). Cases of prolonged carrying at Bossou occurred during the dry season [
21], but prolonged carries in Mahale happened throughout the year [
27]. In sum, we found no clear support for the hypothesis that carrying duration varied according to within-site environmental conditions.
The unawareness of death hypothesis concerns what mothers may or may not understand about death [
24,
25]. One prediction arising from this hypothesis is that younger and/or first-time mothers carry dead infants for longer in order to gather more information and be certain the infant is dead. None of our candidate models that included either firstborn status or mother's age at death performed significantly better than the null model. In fact, aside from one outlier (APb1), mothers tended to carry firstborn offspring for shorter durations than later born offspring (though this parameter estimate included 0 in the 95% CI). Given that we only have nine firstborn offspring in the dataset of observed corpses, more cases are needed to confirm this pattern. In the above-mentioned anthropoid primate meta-analysis [
17], younger females exhibited shorter durations of carrying (regardless of firstborn status) in contrast to prediction. However, because datasets from across 18 different species were combined, mother's age was partitioned into two categories, younger versus older (according to median life span for the species). That difference in methodology, coupled with the smaller sample of nine chimpanzees in that study, may account for our different findings. These authors also found that infants that died of sickness or were stillborn were carried for longer than those that died of infanticide or unnatural causes (such as electrocution) and suggested that more violent/obvious causes of death resulted in faster detection of death and abandonment of corpses. By contrast, we found that candidate models including cause of death provided no better fit than the null model.
Our examination of the specific types of behaviour directed at the corpse suggests that mothers/primary carriers rapidly recognized a change of state in the infant, given that atypical carrying postures were exhibited soon after death. In fact, these postures more closely resembled those used for objects than live infants [
53,
54] and would be uncomfortable or harmful for a live infant, suggesting that mothers persisted in infant corpse carrying for hours or days after recognizing a change in state. There are also multiple sensory death cues: failure to respond to tactile stimulation, olfactory cues of putrefaction, and the presence of flies. While we do not have consistent quantitative measures of distance for our cases, observers often described a process in which the mother gradually increased her distance from the corpse, as was reported in [
29]. As the post-mortem period progressed, siblings and others were permitted to interact with the corpse in a manner similar to object play, or roughly handle the corpse. In very rare cases, mothers exhibited rough handling or ate parts of the corpse themselves. These qualitative reports and our model comparison results do not support the unawareness of death hypothesis.
With regards to the broader question of whether there is evidence of a human-like death concept in chimpanzees, our data provide additional evidence that chimpanzees understand at least two of the four subcomponents reviewed in [
2]: non-functionality, given the changes in behaviour described above, and irreversibility, given the eventual abandonment of the corpse. Whether or not chimpanzees understand death as universal is not possible to examine with our data, and indeed, is difficult to explore without a common language. The causality component is the last to be acquired by human children and its importance is debated given its reliance on knowledge of biological processes that have only come about through modern science and medicine (A Gonçalves 2019, personal communication). Das
et al. [
17] argued that anthropoid primates comprehend causality, because infants that died of unnatural and externally observable causes were carried for shorter periods than those that died of natural causes. However, we did not find that cause of death explained the variation in duration of infant carrying in our larger, single-species sample.
Evaluating our findings according to the three-level model of death awareness proposed by Gonçalves & Carvahlo [
3], our data provide support for the first two and suggest that chimpanzees quickly distinguish between animate and inanimate and also integrate sensory and contextual cues to discriminate between living and dead. The third level of death awareness encompasses the four subcomponents of a human-like death concept described above, and we have provided additional evidence in accordance with [
2] for chimpanzees' understanding of non-functionality and irreversibility. Evidence for the subcomponents of universality and causality will probably need to come from rigorous and ethical cognitive experiments conducted in captive settings [
19].
Given that we found no evidence for the main existing hypotheses for ICC, the question remains as to why chimpanzee mothers, if they have distinguished between alive and dead, continue to carry their infant corpses for multiple days after death and often exhibit caretaking behaviours such as grooming. This behaviour could be interpreted as lack of awareness that death has occurred, but the behavioural changes we described above suggest otherwise. An intriguing possibility is that infant corpse carrying represents a primate analogue of human grief [
17,
55]. However, grief has been difficult to systematically define and operationalize; even in humans, grief is recognized as a ‘highly individualized and dynamic process’ [
56, p. 119]. Evidence in non-verbal animals must come from behaviour, and an increasing number of species have been reported to exhibit behaviours similar to humans. These include prolonged association with the corpse [
8,
57–
59], decreased appetite [
12,
17,
57,
60], and social isolation and/or avoidance [
17,
22,
57]. How non-human animal emotions are defined, measured and whether they differ in degree or kind from human emotions is an active area of scientific debate (see [
61,
62]). Responses to death may be a particularly fruitful area of focus for the study of primate emotions.
Much work remains to be done to achieve a more complete understanding of non-human animal responses to death. We echo the calls of our colleagues [
3,
16] for more detailed descriptions of cases using standardized terminology and variables, and collection of physiological data such as stress metabolites. Detailed analyses of changes in maternal activities budgets, such as reduced feeding and socializing, as well as others' consolation behaviours towards mothers, will help to provide a more complete picture of a species' understanding of death. These data are extraordinarily difficult to collect and accumulate given the rarity of observed deaths in the wild, which highlights the importance of long-term studies for understanding of the complexity of animal behaviour. While many questions remain, observations reported here and elsewhere challenge Heidegger's [
1] assertion that ‘Animals cannot do this'.
Posts
