Asian elephants (Elephas maximus) reassure others in distress

(a) Study area and subjects

This study was conducted at the Elephant Nature Park (the “Park”) in the Mae Tang district of Chiang Mai province, Thailand. Although the Park owns many of the elephants on-site, some are leased or contracted so that the general elephant population changed regularly during the study period. The data in this study refer to 26 elephants with approximate ages ranging from 3 to 60 years old, although due to unverifiable records, ageing elephants precisely was impossible. Although genetic tests on the relatedness of the elephants were never done, it is reasonable to conclude based on the relayed life histories of the individual elephants that all individuals, except for mother-juvenile pairs brought to the Park together, were unrelated. Each elephant was taken care of by one or two mahouts (elephant caretakers) every day. Adult male elephants (n = 4) were completely excluded from the study as they were regularly prevented, for safety and husbandry reasons, from participating in most of the natural, social interactions within groups. When a female was first brought to the Park, she was generally allowed to integrate with a smaller group of elephants. In this study, these smaller, social units (generally of n = 5–7 individuals) are labeled “managed groups” because they consisted of individuals that spent most of their social time together under the guidance of their mahouts. There was no single herd at the Park, but six individual managed groups that interacted at specific times during the day. These groups were delineated based on interviews with the Park mahouts during data collection but prior to data analysis.

Each day, elephants followed a specific routine established by Park management. Mahouts moved their elephants to a specific location on the property, as a managed group, beginning at 0700 h. They ate at a central location at 1130 h – fed either by their mahouts or visiting tourists – bathed communally at 1300 h and 1630 h, and returned to their night shelters, in which they were tethered for the night, at 1700 h. Mahouts moved elephants with vocal commands or by grasping their ears or legs and walking them to different locations on the property. Throughout the day, elephants were left to graze or play in various parts of the property within their social groups. Although individual elephants were generally allowed to interact with members of other managed groups, the mahouts often intervened at unpredictable times to separate volatile pairings.

(b) Defining distress

Because there is very little literature on Asian elephant behavior in general (but see Sukumar, 2003; Sukumar, 2006; de Silva, Ranjeewa & Kryazhimskiy, 2011), the more detailed literature on African elephant behavior (Loxodonta genus – e.g., Douglas-Hamilton & Douglas-Hamilton, 1975; Moss, 1988; Poole, 1996; Payne, 2003) is often applied to Asian elephants as well because of their relatively close phylogenetic proximity (Payne, 2003). Douglas-Hamilton & Douglas-Hamilton (1975) and Lee (1987) describe distress in individual elephants, specifically infants, based on specific vocalizations and stimuli. Infants give a specific call – either an infant roar or squeal – and assume an alert posture where the head is raised, the ears are extended, the tail is raised and the trunk is either raised or stiffened outward (Olson, 2004). Roars, rumbles, and trumpets are often given in response to infant distress calls, or as a signal of an adult’s own distress. Using (1) Lee’s (1987) definition of distress events in calves as those that result in “a dramatic response on the part of other animals... rushing to assist the calf” (p. 287), (2) Bates et al.’s (2008) definition of empathic responses to distress as: “A voluntary, active response to another individual’s current or imminent distress or danger, that actually or potentially reduces that distress or danger” (p. 208), and (3) a comprehensive ethogram of elephant behavior with specific attention to those behaviors occurring when an infant or adult is distressed or agitated (adapted and expanded from Olson, 2004), we define a distress event in elephants as follows:

A distress event is one resulting from an unseen or seen negative stimulus (e.g., mahout-driven separation or movement of individuals, conspecific intimidation or aggression, group separation, environmental threat or accident) that causes an individual to become agitated and to signal such agitation to others (which can be visually identified with specific changes in body state – ears forward, tail erect – and movement, and acoustically identified by various vocalizations, specifically trumpets, roars and rumbles).

(c) General data collection

We chose locations on the property from which to collect data to ensure both a full view of pre-selected managed groups and the observer’s safety. These locations included viewing platforms constructed specifically for observations, and in fields in close proximity to mahouts. Observation locations were chosen based on three factors in decreasing priority: (1) safety of observation vantage point at any given time, (2) view of a maximum number of managed groups at the beginning of the observation period, and (3) view of the managed groups from which there were the least amount of data. The property was approximately 55 acres in total size, but only 30 acres were observable for this study. The property was divided into four grids for observation purposes, and an observation location was chosen within a grid based on the aforementioned factors.

On average, data were collected during 1–2 week periods each month from April, 2008–February, 2009. General observation periods ran for no less than 30 min and no more than 180 min per session from 0730 to 1030 and from 1400 to 1630, with scan samples taken every 10 min. Data on proximity distance only were collected for relationship quality within elephant groups. All observation periods began after 10 min of no mahout interference on elephant behavior, and individual scan samples were cancelled if such interference occurred within a given 10-min period. All-occurrence sampling was used for distress behaviors and the reactions of others to these behaviors (Altmann, 1974). In addition, if an interaction was clearly and completely observed outside these specific observation periods, the same data were collected ad-libitum (<20% of total cases), and a subsequent control observation period (see below) was scheduled.

(d) Post-distress data collection – PD and MC observations

Although the human staff responsible for the elephants’ care artificially constructed the managed group over several years, we focused on spontaneous, affiliative behavior reflective of natural, social interactions (de Waal, 1982; Sukumar, 2003, 2006; de Silva, Ranjeewa & Kryazhimskiy, 2011). Post-distress data for this study were collected at the Park on 26 semi-free ranging individuals in six managed groups following the PC (post-conflict, or PD, post-distress)/MC (matched-control) methodology developed for reconciliation and consolation behavior in primates (de Waal & Yoshihara, 1983).

The PD period was an observation period in which all approach and affiliative behavior was recorded (as were all data on potential stimuli for distress, individuals present within 50 m, and date, time and weather), for a 10-min block following the first distress display. We chose a 10-min duration because it (a) follows the methodology employed with many other non-human species (see Aureli & de Waal, 2000, for a review), and (b) far exceeded the average time of first bystander response to another’s distress in a baseline observation period conducted by the first author prior to the start of data collection. Because elephant interactions may involve multiple distressed individuals (Lee, 1987), the first individual to vocalize, or display a distress behavior was labeled the victim and thus the focal individual in each PD period. If more than one individual responded simultaneously, the rarest case (if known, the least-often distressed individual) was chosen for observation. Each PD period was compared to an MC (matched control) period, or another 10-min block of observation taken of the victim and bystanders on the next possible day following the PD. An MC period was selected when as many variables from the PD – in prioritized order: high percentage of original individuals present, location, time of day and weather – could be maintained, and, most importantly, no new distress event occurred in the 30 min prior to (or during) the period of observation. An MC was collected within seven days of its corresponding PD (in 80% of PD/MC cases, the MC was collected within 48 h). If an MC was conducted when an elephant that had made contact with the distressed individual in the corresponding PD was absent or more than 25 m away, that PD/MC dyad was excluded from the analysis to avoid biasing the data in favor of our predictions.

(e) Scan-sampling for proximity – “friends” and “non-friends”

We attempted to differentiate between contact directed toward “friends” (closely-bonded individuals) and “non-friends” (those outside managed groups) by collecting 68 h of scan-sampling proximity data (for procedure, see Romero & de Waal, 2010). Although mahouts did not interfere with most day-to-day social interactions within established, managed groups (and thus we were able to specify controlled parameters for the PD/MC data), they often discriminately prevented outsider elephants from coming too close to avoid potential conflict. Such conflict between elephants at the Park was also not representative of natural, wild elephant groups (in which conflict is relatively rare), probably due to a high level of unrelatedness within and between managed groups at the Park. Unfortunately, we were forced to exclude the scan-sampling data from our analysis due to circumstances beyond our control. Thus, we were unable to measure relationship quality and its effect on levels of affiliative behavior in this study.

(f) Analysis

We used Wilcoxon signed-ranks tests (two-tailed) to analyze the differences between PD and MC pairs because of the relatively small sample size. The data were analyzed by focal individual to avoid biasing the data toward any particularly well-represented focal elephant. In addition, the McNemar test was used to assess the presence or absence of elephant bunching behavior within PD/MC observations (Siegel & Castellan, 1988). All tests were two-tailed, and P-values were compared to an alpha level of α = 0.05.

(a) Physical affiliation following distress

To assess reassurance behavior, we first recorded the timing of the first affiliative interaction between the victim (the first individual in a group to display distress behavior, i.e., vocalizations and body state changes signaling emotional distress or agitation) and any bystander(s), with physical contact and affiliative vocalizations analyzed separately. These data were collected during the 10-min PD period and then compared to the timing of the first affiliative interaction in the corresponding MC period. Following standard procedures developed in primate studies (e.g., de Waal & van Roosmalen, 1979; de Waal & Yoshihara, 1983; de Waal & Aureli, 1996; Romero & de Waal, 2010), PD/MC pairings were split into three categories: attracted (pairings in which the first affiliative contact occurred earlier in the PD than in the MC, or no contact occurred in the MC following contact in the PD), dispersed (contact occurred earlier in the MC than in the PD or not at all in the PD), and neutral (affiliative contact times did not differ in the PD and its corresponding MC, or no contact occurred in either) (e.g., de Waal & Yoshihara, 1983; Veenema, Das & Aureli, 1994; Verbeek & de Waal, 1997; Aureli & de Waal, 2000).

There were 84 PD/MC observations (and thus 84 distinct initial instances of distress signals) across 18 different focal individuals (mean number of PD/MC observations per individual = 9.5, range = 1–38). Within the 84 PD/MC observations, there were a total of 183 focal-bystander dyads, 171 of which involved at least one affiliative physical contact (e.g., Fig. 1) during the PD period (93.4%). 53 of the 84 PD/MC observations included affiliative contact by multiple individuals directed toward a single focal individual. 12 of the 84 observations were the result of an identifiable stimulus for distress – either directed aggression or a feature in the environment (e.g., helicopter, human or dog in close proximity) – that caused distress first in a single individual. The sample size did not allow for further analysis by stimulus type. In our analysis of affiliative contacts, we were concerned only with the first contacts between bystanders and the focal individual in each of the 84 PD/MC observations. The majority of affiliative contacts occurred within the first min following distress (Fig. 2; see Movie S1 for an example of affiliative contact), and a Wilcoxon signed-ranks test performed on the data by focal individual showed that the difference in frequency of these contacts per individual subject in the first min of the PD (mean ± SD = 7.50 ± 8.49) versus the MC (mean ± SD = 0.44 ± 0.86) was significant (Z = 3.56, n = 18, P < 0.001).

We categorized attracted and dispersed pairs based on whether or not each interaction was “solicited” (the focal, distressed individual approached a bystander to seek reassurance) or “unsolicited” (a bystander was the first to approach the focal, which is sometimes called “true consolation” in primate studies – Call, Aureli & de Waal, 2002; Koski & Sterck, 2007). When the first affiliative contacts between the focal individual and bystanders in each of the 84 PD/MC observations were analyzed (the usual first step in assessing consolation data – e.g., de Waal & van Roosmalen, 1979; de Waal & Yoshihara, 1983; de Waal & Aureli, 1996; Romero & de Waal, 2010), a significant difference was found between the proportion of attracted and dispersed pairs in both unsolicited (Z = 3.31, n = 18, P < 0.001) and solicited contacts (Z = 2.69, n = 18, P = 0.007; Table 1). Across the 18 focal individuals, unsolicited contacts (mean ± SD = 8.83 ± 11.93) occurred significantly more often than solicited contacts (mean ± SD = 1.33 ± 1.71; Z = 2.47, n = 18, P = 0.014). The two most prevalent types of physical contact given by bystanders were trunk touches to another individual’s genitals (38.6% of touches), and trunk touches around or inside another’s mouth (35.1%; Fig. 3).

(b) Vocal affiliation following distress

Because elephants primarily use acoustic modalities for communication (e.g., Poole, 1996; Payne, 2003; Nair et al., 2009; de Silva, 2010), we also looked at bystanders’ vocalizations in response to distressed individuals. In a comparison of first bystander vocalizations in the PD and MC periods, we found that bystanders vocalized earlier following distress than in control periods in a significant number of PD/MC observations (proportion of attracted pairs: mean ± SD = 97.11% ± 8.81%; dispersed pairs: 2.22% ± 8.61%) across 18 focal individuals (incidentally, only three of these focal individuals never had a bystander vocalize when they were distressed: Z = 3.42, N =18, P < 0.001). Bystander elephants most often chirped (a vocalization often emitted when individuals are in close-proximity to one another – 31.8% of vocalizations) or audibly trunk-bounced (interpreted as a sign of agitation or distress – 24.7% of vocalizations) following distress signals from the focal animal (Olson, 2004; Nair et al., 2009; de Silva, 2010, see Fig. 3).

In addition, we assessed differences in the behavior of bystanders in relation to the behavior of distressed individuals between PD and MC periods. Vocalizations may signal agitation or excitement in elephants and are usually paired with similarly functioning physical and postural displays (Olson, 2004; Nair et al., 2009; de Silva, 2010). Bystanders adopted the agitated behavior of the originally distressed focal individual in the PD (e.g., ears presented forward with an erect tail, usually followed by several vocalizations and sometimes with simultaneous urination and defecation – Olson, 2004; Bates et al., 2008), yet showed no such signs of agitation or distress in the MC in 157 of the 171 dyads in which physical contact occurred (91.8%: mean ± SD = 8.72 ± 9.51 dyads per focal individual; Z = 3.56, n = 18, P < 0.001).

(c) Behavior among bystanders

The previous results refer to contact by bystanders to a distressed, focal individual, but we also analyzed contact between bystanders in PDs in which there were multiple individuals present. Bystander-bystander physical contact occurred in 37 of the 84 PD periods, and, like in victim-bystander contacts, occurred earlier following the victim’s distress in the PD period than in the MC in a significant number of interactions across 19 possible bystanders (proportion of attracted pairs: mean ± SD = 97.37% ± 8.36%; dispersed pairs: 2.63% ± 8.36%; Z = 3.85, n = 19, P < 0.001).

Elephants may quickly form a close circle, known as “bunching,” around their young in anti-predator defense (e.g., Moss, 1988; Poole, 1996; McComb et al., 2001; Bates et al., 2008). Bunching involves the coming together of multiple individuals around the distressed elephant so that all individuals are within trunk’s reach of one another (Nair et al., 2009). To systematically assess whether individual signs of distress trigger such behavior, we looked at the occurrence/non-occurrence of bunching in PD/MC observations. We excluded all observations in which less than four individuals were present (this excluded n = 7 focal individuals altogether). In 30 of the 42 qualifying PD/MC observations, bunching around both juveniles and other adults occurred following distress and never in the corresponding control periods (McNemar change test comparing presence or absence of bunching in PD and control periods: χ² = 28.03, df = 1, P < 0.001).