By interacting with the environment, organisms develop ways to avoid dangers and recognize predators or non-mutualistic species such as parasites (Deecke et al., 2002; Marzluff et al., 2010; Papworth et al., 2013). Accruing the ability to collect information from the ecosystem and other peers ultimately increases the species’ fitness (Tuttle and Ryan, 1981). Encoding individual vocal identity should be adaptive for species above all living in large and complex societies (Rhebergen et al., 2015). For instance, from a distance, herbivores can identify the scents and sounds from predators and flee them before they attack; or birds recognize and react to alarm calls from other prey species to avoid predators (Fallow and Magrath, 2010).

With the impinging development of our society, animals interact with humans more and more frequently. Different people may pose different levels of threats (or benefits) to inhabiting animals (Lee et al., 2011; Chen et al., 2016). This is also the case with animal husbandry, where domesticated animals interact frequently with handlers. Here domestication is defined as evolutionary process forced by human influence (Price, 1984). Domestication selection may have therefore allowed the formation of artificial forces selecting in favor of animals able to discriminate their owners from unfamiliar individuals (Topál et al., 2005). Consequently, differentiating threatening humans from beneficial ones could have become a requisite for both domesticated and wild animals to cope in the “Anthropocene” (Bond et al., 2020).

Many kinds of research on discrimination ability of animals on human individuals used domesticated animals as study subjects, such as the domestic cat (Felis catus; using acoustic cues; Saito and Shinozuka, 2013), the dog (Canis familiaris; using visual cues and acoustic cues; Adachi et al., 2007), the pig (Sus scrofa domesticus; using visual signals and olfactory signals; Koba and Tanida, 2001), the cattle (Bos taurus; using individual humans as cues; Taylor and Davis, 1998), and the lamb (Ovis aries; using individual humans as cues; Boivin et al., 1997). As a result, some argue that artificial selection occurring during domestication could be an evolutionary force influencing animals’ ability to discriminate among human individuals (Topál et al., 2005). However, other studies reveal that some wild animals show some ability to recognize humans on an individual level, and that this ability develops by merely interacting with human beings. For instance, the wild magpie (Pica pica) is aggressive only toward those specific humans who pose a threat to their nests (using human individuals as a cue; Lee et al., 2011). Mockingbirds (Mimus polyglottos; Levey et al., 2009) and American crows (Corvus brachyrhynchos; Marzluff et al., 2010) display similar discrimination abilities. These studies in domesticated and non-domesticated animals indicate that they can learn and apply their knowledge to discriminate against different human individuals while interacting with them. Instead of domestication, the mere proximity to humans could be enough to develop discrimination ability on human individuals for these animals.

This type of research customarily resorts to visual stimuli, such as the images of humans or their behaviors within a short distance (Levey et al., 2009; Marzluff et al., 2010; Papworth et al., 2013). To avoid dangers, some species could use more efficient ways such as acoustic cues (i.e., human voices) to assess threats at an extended range, such as from humans out of sight (Fallow and Magrath, 2010). However, only a few studies have explored if non-domesticated animals can recognize human individuals using these acoustic cues (carrion crow, Corvus corone; Wascher et al., 2012; cheetah, Acinonyx jubatus; Leroux et al., 2018). Comparative research on phylogenetically distant species can promote the further understanding of evolutionary drivers on the animals’ discrimination ability.

Elephants are social animals with advanced cognitive abilities to group living and group cooperation (self-recognition, Plotnik et al., 2006; empathy, Plotnik and de Waal, 2014; cooperation, Plotnik et al., 2011; Li et al., 2021). Furthermore, elephants excel at learning and at reproducing acoustic cues (Heffner and Heffner, 1982; Payne et al., 1986; McComb et al., 2003; Poole et al., 2005). For instance, it was reported that Asian elephants (Elephas maximus) can imitate human speech so accurately that local people cannot tell apart the animal imitation from human voices (Stoeger et al., 2012). The Asian elephant can use various vocalizations for communication between social groups and within the same group, among group members, including trumpets, chirps, roars, and rumbles of different types (Nair et al., 2009). And when perceiving disturbance, Asian elephants can modulate their vocalizations to alert conspecifics (Sharma et al., 2020). Humans have a long history of taming Asian elephants for logging and the carrying of goods in some countries, such as Myanmar and Thailand (Sukumar, 2003). During these long-lasting interactions, it has been reported that the Asian elephant is able to understand some human behavioral cues (Ketchaisri et al., 2019).

Furthermore, with humans encroaching and fragmenting the natural habitats, and with land-use change, there is an increasing overlapped distribution and interaction between humans and elephants (White and Ward, 2011; Liu et al., 2017). Wild elephants near human residences often raid crops, destroy properties, and can sometimes cause human casualties (Chen et al., 2016). In this case, different residents may apply different methods to keep elephants from human settlements (Aziz et al., 2016; Wijayagunawardane et al., 2016). During human-elephant interactions, elephants may develop the ability to discriminate between threatening individuals to non-threatening individuals, and they have excellent memory capacity (Rensch, 1957; Hart et al., 2008). It is valuable to test if Asian elephants can discriminate against humans on an individual level.

Research using two Asian elephants as study subjects at a zoo revealed that elephants showed no significant behavioral difference after hearing acoustic playback of familiar and unfamiliar humans (Polla et al., 2018). However, Asian elephants mainly live in forest habitats, where acoustic signals could be critical for detecting danger. For instance, a previous study shows that Asian elephants can discriminate the tiger growl (threatening to elephants) from the leopard growl (not threatening to elephants; Thuppil and Coss, 2013). Therefore, we predict Asian elephants to be good at acoustic recognition, and the results of the previous study might be consequential to the small sample size and the housing conditions.

To explore if the Asian elephant can recognize familiar humans on an individual level by acoustic cues alone, we studied semi-captive Asian elephants (semi-captive here is defined as captive elephants that can range and forage freely in forest) in Myanmar. In the Indian subcontinent and Southeast Asia, mahouts (or “oozies”; i.e., whose job is to take care of elephant exclusively) used to work with these animals for a living. Today, mahouts continue to be solely in charge of the daily care of the elephants. In particular, we tested if these semi-captive animals could discriminate between their own mahouts and other mahouts (unfamiliar individuals), when there was only an acoustic stimulus given to them. As mentioned, the aim was to test their ability to discriminate between familiar and non-familiar mahouts. We predicted that Asian elephants could be able to recognize the voice from their own mahouts in an accurate manner.

The elephants’ age in this study ranged from 7 to 36 years (Table 1). Ten of them were captive-born, and eight of them were wild-caught. On average, the time our mahouts spent with the elephants ranged from 6 months to 16 years. Overall, the elephants responded more to the command from their own mahouts (familiar humans), regardless of whether they were captive-born or wild-caught (Figure 2; see parameters of Model 1 in Table 2). The female elephants responded more than the males to the commands, and the younger elephants responded more than the old elephants. Most importantly, we found that with more extended proximity (care time from the mahouts), the elephants showed more active responses (Figure 2; see parameters of Model 1 in Table 2).

The average successful response (i.e., the elephant responded only to the matching mahout and did not respond to the commands from non-matching mahouts) rate was 78.8% (i.e., the total number of successful responses/numbers of total tests). Among all the 18 elephants, one elephant called NNN, a female elephant, responded 100% correctly to her own mahout and never responded to the commands from non-matching mahouts (see Table 3 for the performance of each elephant in the test). No matter whether the matching mahout or non-matching mahouts gave commands, the female elephants showed significantly more active responses than the male elephants (Figure 3; see parameters of Model 2 and Model 3 in Table 2).

When it was the matching mahout giving out the command, the elephants who had lived longer together with their mahouts (i.e., longer care time of their mahouts, or longer proximity with mahouts in general) showed significantly more active responses toward the commands from the matching mahouts (Figure 4; see parameters of Model 2 in Table 2). Younger elephants showed significantly more active responses to matching mahouts than older elephants when it was the matching mahout giving out the command. Yet, when non-matching mahouts were giving the commands, only the sex of the elephants had an impact on their responses, with the females responding more often than the males (Figure 4, see parameters of Model 3 in Table 2).

The response duration of the elephants when their own mahouts gave out the commands revealed similar patterns (Figure 4). Overall, whether the elephants were captive-born or wild-caught had no influence on the elephants’ response duration. Here, female elephants responded more quickly (i.e., shorter response duration) than male elephants (Figure 4; see parameters of Model 4 in Table 2). Most importantly, the elephants responded faster when their mahouts spent more time with them in the past (i.e., care time of mahouts or proximity with mahouts; Figure 4; see parameters of Model 4 in Table 2). Furthermore, younger elephants responded more quickly than older elephants when it was the matching mahout commanding (Figure 4; see parameters of Model 4 in Table 2).

Our study finds out that Asian elephants can unambiguously recognize the commands from individual mahouts by acoustic cues (Figure 2). Asian elephants who have been taken care for an extended time can develop a clear ability to discriminate between familiar and unfamiliar humans, regardless of whether they are captive-born or wild-caught (Figures 2, 4). Although ten of the semi-captive elephants in this study are captive-born, they might not go through an evolutionary process of domestication. We therefore argue that, at least for the Asian elephant, domestication is not instrumental in developing this discriminatory ability, and that possessing a familiarity with their handlers is sufficient.

Polla et al. (2018) reported that Asian elephants did not show any significant difference in their trunk-reaching frequency and bout duration when triggered by the voices of familiar and unfamiliar humans. However, with a small sample size of two zoo-housed elephant individuals tested over a limited number of human voices, the auditory discrimination rate they report is hardly conclusive (Jennions and Møller, 2003; Button et al., 2013). Our study used 18 elephants to look for a more general pattern. Furthermore, we used a command pertaining to locomotion, which implied an objective judgment of the animals.

In our discrimination test, the female elephants showed more frequent and faster responses than the male elephants (Figure 4). This may be attributable to the fact that female elephants live in social groups for all their lives (Vidya and Sukumar, 2005; Nandini et al., 2017). This was indeed the case of the elephants tested here. During this time in the elephants’ lives, the females are expected to develop more cooperative behaviors than the male elephants (such as taking care of each other’s calves; Sukumar, 2003). In addition, the younger elephants responded more frequently and faster to their matching mahouts, which may imply that younger individuals have better learning abilities than older elephants (Mader and Price, 1980; Head et al., 1995) or that older elephants are habituated to the commands from their matching mahouts. It is less likely that this pattern is caused by younger elephants being more docile or tamed than older ones, in which case the younger elephants may have responded more even when it was up to the non-matching mahout to give out the command. Yet, younger elephants and older elephants perform the same when they hear the command from unfamiliar mahouts (Figure 4). Further research is needed before drawing conclusions about this age factor. Most importantly, when mahouts spent more time with the matching elephants in the past (longer care time), elephants responded more and faster to the matching mahouts (Figure 4). This result indicates that the human acoustic cue discrimination may be a learning process, an adaptation to the extended exposure and received stimuli from humans become very familiar.

As our study assigned three elephants to a group, there might be a behavioral dependency effect between them. Yet, during the experiment, we only recorded 7 interaction events between elephants (playing with each other; i.e., using trunk to push or tangle with each other), and 28 events of one elephant following another elephant to the finish line, out of 351 trials. Most of the time, they stood still in the initial place or engaged in recreation with sand most of the time. In addition, because of equipment and electricity limitation in the elephant camp, we assigned three mahouts in a group, to avoid elephants using olfactory cues to discriminate among mahouts. Yet a better experiment design using playback experiment is strongly recommended in the future study, to remove the effect of olfactory cues.

Like for all the other wild megafauna, the main threat for Asian elephants is now from human beings (Desai and Riddle, 2015). Human-elephant interactions frequently happen in elephant-range countries (Chen et al., 2016), where local residents use firecrackers or other sounds, like the playback of felid growls, to deter elephants from raiding their crops (Thuppil and Coss, 2016). However, despite this artificial change in the acoustical structure of the animals’ environment, elephants have given proof to quickly adapt to these harmless dissuasive techniques (Joshi, 2013). With large brains, Asian elephants can adapt to degraded environments (Sol et al., 2008), by remembering the link between sounds and context. In Kenya, using acoustic cues only, wild African elephants can recognize subgroups of people from different ethnicities who pose threats or no threats to them (McComb et al., 2014). Our study provides evidence that semi-captive Asian elephants can recognize human voices at an individual level accurately. These elements may point toward why elephants are adapting to living near human residences, making human-elephant conflicts a wicked problem. By living in group and interacting with humans, social learning abilities of elephants may have likely promoted the ability to discriminate among humans both on a subgroup and on an individual level.

To conclude, this study demonstrates that the Asian elephant is able to discriminate between familiar and unfamiliar humans by means of acoustic cues alone. Females and younger individuals perform better in the discrimination test. The longer the elephants have interacted with the matching mahouts, the better they perform in human discrimination. Further comparative research is needed for the assessment of this ability in other non-domesticated animals, to understand the adaptation and evolutionary drivers of novel, functional skills under human-dominant circumstances.

The raw data supporting the conclusions of this article will be made available by the authors, without undue reservation.

The animal study was reviewed and approved by the Biomedical Ethics Expert Committee of Xishuangbanna Tropical Botanical Garden, Chinese Academy of Sciences (protocol no. XTBG-2020-11).

L-LL, R-CQ, and RP designed the study and wrote the manuscript. L-LL conducted the experiments. RH and L-LL analyzed the data and conducted graphing. All authors read and approved the final manuscript.