2.1 Subjects and Housing

This study took place in the Serengeti-Park Hodenhagen, Germany, in 2022. Two L. africana cows (mother and daughter), Bibi (~36 years) and Panya (14.75 years), currently living in the safari-park/zoo participated in the present experiment. Three other females and two males were living together in their herd. The animals were kept in protected contact. Before their current location, Bibi and Panya, lived in Tierpark Berlin, Germany until November 2008, at which time they were transferred to Zoo Halle, Germany. In November 2017, Bibi was relocated to the Serengeti-Park Hodenhagen. Panya (with her 1-year old male calf Ayo) was then reunited with Bibi in August 2020. The elephants were tested in an outside pre-enclosure individually, but could reach their trunks through the bars towards the experimental setup.

For this experiment, first an establishment phase was conducted for which three current keepers of the Serengeti-Park Hodenhagen were included. For the main experiment, three former keepers from the Tierpark Berlin (last interaction > 13 years), who had worked with the elephants until their move, were used as participants. They had accompanied the animals for 21, 10, and 8 years in the case of Bibi, and for 15 months in the case of Panya. For each keeper, an unfamiliar person was selected as a counterpart in the two-choice object test, this person had never been in one of the zoos where the elephants live/lived and has therefore never had any contact with the elephants. Familiar and unfamiliar people were matched according to approximate age and body stature. None of the persons were related to each other, and all were men.

2.2 General Procedure and Experimental Set-Up

A set-up similar to that of a previous study was used and served as a basis (Polla, Grueter, and Smith 2018). In particular, Polla, Grueter, and Smith (2018) investigated whether two E. maximus cows at the Perth Zoo in Australia could distinguish between their current keepers and people unfamiliar with the elephants in a simultaneous two-choice object test. For this test, only single auditory, visual and olfactory stimuli were used. Here, we examined whether two L. africana cows could also make these differentiations, in addition to distinguishing between their former keepers and people unfamiliar to them.

The experiment took place in the pre-enclosure (186.7 m²) next to the large shared main enclosure (~5500 m²). The set-up itself consisted of two racks (150 cm high), each containing a sound box (Yamaha MS 202 II). Attached to each shelf was an aluminum pop-up poster frame in the front, which rose above the rack by another 80 cm. Also mounted inside the rack, were two hooks for the olfactory stimuli. Each of the two racks served as a holding device for a person's stimulus in the experiment (Figure 1). The set-up was positioned 280 cm apart from the fence, just near enough for Bibi, the larger of the two elephant cows, to reach the set-up with her trunk. Two cameras (GoPro HERO9 Black v01.72.00) were used to record the elephants' behaviors for later analysis. At a distance of approximately 5 m, one camera was located behind the set-up, and the other camera was located approximately 5 m beside the set-up to record all movements of the elephants (Figure 2). The point of view of the camera at the back also later determined the “left” and “right” positions of the experimental set-ups (see Figure 2B).

Each elephant was tested only once a day. The trial days within the individual test sections were randomized. No keepers were in direct visual contact with the elephant during the test. The area around the elephant enclosure was closed to visitors during the trials, to avoid any distraction.

2.3 Trial Sections

The trial series started in May 2022 and ended in June 2022. A total of 35 trial days were conducted consecutively (Table 1). The start time of the trials was between 9:00 and 10:30 a.m. and was adjusted according to the keepers' work schedule. In the experiment, each animal was alone in the pre-enclosure for 4 min and was free to decide whether it interacted with the experimental set-up or not.

To familiarize the animals with the experimental set-up and routine, an acclimatization period was carried out before the actual experiment. For this purpose, the planned routine without stimuli was conducted on 8 days.

For the establishment, three current keepers and three random people (unfamiliar), were asked to provide each of the three different stimuli. The stimuli were randomized as well as the arrangement of the test persons on the test set-ups (right and left).

Right after the establishment, the trial days of the experiment started, in which stimuli from the former keepers and new unfamiliar people were used. The trial days and the arrangement of the test persons were also randomized. The experiment was then repeated with the same elephants again to obtain a higher number of trails. The order of the trial days and the arrangement of the test persons were again randomized.

2.4 Data Analysis

The recordings of the trial days were randomized and analyzed blindly via the software “BORIS” (Behavioral Observation Research Interactive Software) v7.13.6 (Friard and Gamba 2016). Three parameters were determined for the quantitative statistical analysis. The most important parameter is the “trunk reach duration” (TRD), which describes the time the elephant directs the tip of its trunk towards the stimulus. Linked to this is the “trunk reach frequency” (TRF), that is, how often the elephant points the tip of its trunk at the stimulus. The TRD and TRF were measured whenever the tip of the trunk was 90 cm from the base of the trunk, 70 cm above the ground and pointed directly at the stimulus. The third parameter is the “time of interest” (TOI), that is, the time the elephant is facing a stimulus. The TOI was measured from the time the elephant crossed the bars of the fence until the time its head was behind the bars again. In addition, the initial interest in responses to stimuli by the elephants was recorded, that is, the first TRD/TRF. Due to the temporally different appearance of the auditory stimulus, the initial interest could not be evaluated for these stimuli.

2.5 Statistical Analysis

The statistical analysis and graphs were done with R v4.3.1 (R Core Team 2023). All statistical tests were preceded by a check for normal distribution using the Shapiro-Wilk test. None of the data were normally distributed, so non-parametric tests were used. For the entire study, permutation tests using a for-loop were applied. For each sensory stimulus (auditory, olfactory, and visual) and parameter (TOI, TRD, and TRF), the distribution of the mean differences between the groups (familiar and unfamiliar) was determined. The values within each group were permuted and the significance level was then determined from the observed and expected differences (an example r code can be found in File S4). For each test, 10,000 permutations were performed. The significance level was set at α = 0.05.

During acclimatization, it was determined whether the animals showed a side preference to the experimental set-ups. For this purpose, TRD, TRF and TOI for right and left were tested for statistical significance. If the null hypothesis—no preference for one side—was not rejected for both animals, it could be assumed for further experiments that both animals had no preference and their data from the following phases could be pooled as one data set. For the establishment phase, it was tested if there was a difference in behavior towards familiar (current keepers) and unfamiliar individuals. Since the three parameters were constant for all stimuli and, assuming that the elephants could in principal identify familiar people through all three senses, the data from the three different stimuli were analyzed together. For the main experiment, it was tested for a difference between behavior towards familiar (former keepers) and unfamiliar people, as in the establishment. The two data sets from Experiment 1 and Experiment 2 were combined resulting in $\mathit{n}=12$ trials. Again, both the individual stimuli alone and all three stimuli together were analyzed.

3.1 Acclimatization

Both elephant cows were given 8 days of acclimatization with 4 min each day to get used to the experimental set-ups and the slightly different morning routine. Neither of the two elephants showed a statistically significant preference for one side for any of the three parameters (permutation test, $\mathit{n}=8$. TRD: $p_{\mathrm{Bibi}}=0.204$, $p_{\mathrm{Panya}}=1$; TRF: $p_{\mathrm{Bibi}}=0.559$, $p_{\mathrm{Panya}}=1$; TOI: $p_{\mathrm{Bibi}}=0.994$, $p_{\mathrm{Panya}}=1$).

3.2 Establishment

The establishment phase was first carried out with samples from the three current keepers and three people unfamiliar to the animals. Since neither of the two elephants showed a side preference for an experimental setup, the results from the establishment of both elephants were combined. For both the individual stimuli and all stimuli combined, no statistically significant difference in behavior was found for any parameter for the different groups of people, except for the visual stimuli where the animals showed significantly more TRD and TRF for the familiar stimuli (Table 3). The average TRD for the visual familiar stimuli was 21.9 times as long (${\mathbf{⌀}}_{\mathrm{Familiar}}=1.22\pm 1.30\mathrm{s}$, ${\mathbf{⌀}}_{\mathrm{Unfamiliar}}=0.06\pm 0.14\mathrm{s}$) and the average TRF was 7.9 times as often (${\mathbf{⌀}}_{\mathrm{Familiar}}=1.33\pm 0.82$, ${\mathbf{⌀}}_{\mathrm{Unfamiliar}}=0.17\pm 0.41$) as for the unfamiliar stimuli, however, with large standard deviations.

In addition, the animals showed during the trials with visual stimuli interest first in the familiar stimuli four times each, once in the unfamiliar stimuli first, and once Panya showed no TRD/TRF at all. In the case of auditory stimuli, there was a direct inverse correlation of the first interest, four times first for unfamiliar, once for familiar.

3.3 Experiment Trials 1 and 2

The trial phases were carried out with samples from the three former keepers and three new unfamiliar persons. By testing the differences between Trial 1 and Trial 2, no significant difference could be found (Table 2), therefore the data was merged. With the merged data, a significance can only be found for TRD for the olfactory stimuli (Table 3 and Figure 3). The average TRD for the familiar stimuli was 5.9 times as long as for the unfamiliar stimuli (${\mathbf{⌀}}_{\mathrm{Familiar}}=1.39\pm 1.52\mathrm{s}$, ${\mathbf{⌀}}_{\mathrm{Unfamiliar}}=0.24\pm 0.41\mathrm{s}$), however, with large standard deviations.

When all stimuli were examined together, a statistically significant difference could be seen for both TRD and TRF to the familiar versus unfamiliar stimuli (Table 3 and Figure 4). In total, the average TRD for the familiar stimuli was 3.2 times as long (${\mathbf{⌀}}_{\mathrm{Familiar}}=1.17\pm 1.37\mathrm{s}$, ${\mathbf{⌀}}_{\mathrm{Unfamiliar}}=0.36\pm 1.02\mathrm{s}$) and the average for TRF was 2.4 times as often (${\mathbf{⌀}}_{\mathrm{Familiar}}=0.94\pm 1.26$, ${\mathbf{⌀}}_{\mathrm{Unfamiliar}}=0.39\pm 0.69$) as for the unfamiliar stimuli, however, with large standard deviations.

During the trials with olfactory and visual stimuli, the animals showed initial interest in the familiar stimuli six times and twice in the unfamiliar stimuli; four times they showed no interest in any stimuli.

The average results of the test series as well as the results as raw data can be found in the supplementary materials (see Tables S2 and S3).

4.1 General Observations and Differences in the Elephants

Since Panya was with her 5-year-old son during the day and night, she showed clear discomfort at the beginning of the acclimatization phase when she was spatially separated from him. She displayed this discomfort on the first days by vocalizing loudly, flaring her ears, and prancing around wildly. As the days went by, this behavior subsided, and Panya visibly became more relaxed with the situation. On the first day of testing with the auditory stimulus, Panya and Bibi showed imposing behavior, this involved erect ears and short mock attacks to the set-ups, however, also here a habituation set in quickly. This habituation can also be observed in other experiments with elephants and audio speakers (Goodyear and Schulte 2015).

Bibi was sometimes quite fearless to the set-ups, whereas Panya was much more cautious. The keepers described Bibi as an elephant that may be dominant in her interactions. As a mother of five children, she is a very experienced elephant cow, unlike Panya who is only half her age and is probably not yet able to assess such situations as well as her mother. This may lead to Panya spending more time on the set-ups than Bibi and thus having a significantly higher TOI during all phases than her mother (${t⌀}_{\mathrm{Bibi}}=33.496 \mathrm{s}$, ${t⌀}_{\mathrm{Panya}}=73.397 \mathrm{s}$, permutation test: $p=0.008$, $n=35$). Another explanation could be that since Bibi displays stereotypical behavior (head weaving), she may not have enough time to spend on the set-ups. We could not observe a relationship between the stereotypical behavior and a familiar, or unfamiliar person. With the exception of the overturning of the experimental set-up, no other changes in behavior could be observed that were possibly related to a positive or negative attitude of the elephant towards the test persons, such as temporal gland secretions, or signs of a greeting ceremony.

Conducting experiments outdoors is usually not free from environmental influences. This was the case in this experiment due to the structural circumstances in the zoo. The influence of short-term separation on the elephant cows from other herd members cannot be fully evaluated. The bull, for example, distracted the two elephant cows a few times during their trials, but no change in behavior could be observed after the contact and the lost time in the trial had to be accepted. Elephants engaged in other behaviors such as leaning against a grid in the pre-enclosure and eating grass. This happened on the opposite side of the experimental set-up, but also near the set-up itself dedicated directly to the experimental set-up were impacted. However, such behaviors were accepted within the set 4 min time limit. Furthermore, when the elephants ate grass at the experimental set-up, the time was always included in the TOI. It was repeatedly observed that after eating and during the time it took to pull their heads back through the bars, the animals quickly checked the set-up again with their trunk (TRD). We therefore assume that eating grass at the set-up was partly due to the presence of the set-up. Yet, we cannot fully determine whether this had an influence on the decision of the animals about their position to the experimental set-up.

4.2 Perspectives

Elephants have sex-specific lifestyles, with females living in family fission-fusion herds under an older matriarch and males separating from the group at adolescence and living in bachelor groups, often under older bulls, or sometimes as solitary individuals (Evans and Harris 2008; Murphy, Mumby, and Henley 2019). So far, no sex-specific differences have been studied in recognition studies. However, and because of the different lifestyles, bulls should be included in a further study to investigate whether there are sex-specific differences. In addition, the results should be confirmed in a study with a larger sample size in elephants and former keepers. Besides, an adaptation of the experiment with e.g. cross-modal sensory stimuli or the possibility to have the keepers present would be conceivable. When elephants in zoos undergo reunification after years of separation, they exhibited signs of the greeting ceremony that is common in the wild (Hörner et al. 2021). If signs of the greeting ceremony could also be evaluated in the meeting with the former keepers, it would provide further scientific evidence for the special bond between keeper and elephant. In contrast to this positive relationship, a comparison could be made with individuals negatively linked to the animals, such as disliked keepers or veterinarians. For wild L. africana, such emotion-cognition linked recognitions are made between harmless (Kamba, Kenya) and dangerous people (Maasai, Kenya), and this is observed auditorily (McComb et al. 2014), olfactorily and visually (Bates et al. 2007). This distinction is essential for survival, as there could be significant differences between people with different levels of preference and relationships for the elephants in the zoo.