Elephant banana peeling

VIDEO ABSTRACT.

Elephants are large mammals that rely on their trunk to acquire huge quantities of food. Complex muscles 1 and motor control structures 2 mediate dexterous and lateralized trunk behaviors [3][4][5] , but our understanding of elephants' haptic abilities is limited.
Here we describe the banana-peeling behavior of the female Asian elephant Pang Pha at the Berlin Zoo. Like other elephants, Pha consumes green or yellow bananas as a whole. She rejects brown bananas but, unlike other elephants, when on her own she peels yellow-brown bananas. Pha peels faster than humans by a partially stereotyped sequence of behaviors: she breaks the banana, shakes out and collects the pulp, and discards the peel. When yellow-brown bananas are offered to a group of elephants, she changes her behavior and consumes all bananas as a whole with exception of the last banana, which she retains for later peeling. Banana peeling appears to be rare in elephants and none of the other Berlin elephants engage in peeling, raising the question why only Pha peels bananas. Pha was handraised by human caretakers in the Berlin Zoo, who fed her peeled bananas, but never conditioned her to peel them: we suggest she acquired peeling through observational learning from humans. African elephants appear to be able to interpret human pointing gestures 6 and to classify human ethnic groups 7 , but complex humanderived manipulation behaviors like the banana peeling reported here appear to have only rarely been observed.
For most of our peeling experiments, Pha was separated from other elephants (for details see Supplemental experimental procedures in the Supplemental information). When approached with bananas, Pha repeatedly raises her trunk, a putative begging behavior Correspondence shown by all elephants at the Berlin zoo. She takes the banana with the tip of the trunk, holds it and then breaks it against the side of the distal trunk ( Figure 1A). This is followed by shaking the banana until pulp falls out; Pha grabs the pulp with the trunk tip to eat it and discards the peel. Shaking and peeling are repeated until no or little pulp is left inside the peel and leftovers are checked multiple times with the trunk tip (see Video S1 in the Supplemental information).
Initially, Pha's banana consumption patterns appeared random. We would offer Pha bananas for weeks without her peeling a single one. With time, however, we understood that not all bananas are peeled or even eaten by Pha and that her behavior strongly depended on banana ripeness. Accordingly, we grouped each banana into one of fi ve categories: green, green-yellow, yellow, yellow-brown and brown. Peeling behavior varied strongly with banana ripeness. Pha Like other elephants, Pha consumed green to yellow bananas as a whole ( Figure 1C),  2 (4, N = 196) = 121.87, p < 0.001, and often rejected brown bananas ( Figure 1D; Fisher's exact test p < 0.001).
Besides banana ripeness, the social setting shaped Pha's bananapeeling behavior. We fi lmed social banana consumption in Pang Pha, her daughter Anchali and female Asian elephant Drumbo ( Figure 1E). Elephants were instructed to wait for a go signal, while we placed yellow-brown bananas (which Pha would usually peel) in their enclosure. Elephants consumed bananas very fast -one banana every two seconds in the case of Pha. During social feeding Pha consumed most of the yellow-brown bananas as a whole, which differs from her behavior when solitary ( Figure 1F). Pha directly ate the majority (61%) of all the socially fed bananas, whereas she peeled most of them when on her own, and rejections were more frequent in social feeding,  2 (2, N = 99) = 39.81, p < 0.001. Furthermore, there was a signifi cant difference between the last banana, which Pha often retained for later peeling, and the bananas she consumed before (Fisher's exact test p < 0.001). Of the fi rst to second last banana, Pha ate more than two-thirds directly and peeled only one banana, whereas she peeled 60% of the last bananas ( Figure 1G).
Banana peeling seems to be rare in elephants. Anecdotal internet videos provide proof that other elephants peel bananas as well, but we did not observe other elephants at either the zoo in Berlin (n = 4 Asian elephants) or the zoo in Vienna (n = 6 African elephants) to ever peel bananas. The fact that Pang Pha's daughter Anchali did not acquire banana peeling suggests that this skill is not easily transmitted by learning. Pha's tendency to peel only when on her own might have limited possibilities for observational learning. Still, the other elephants presumably watched her peel bananas numerous times. Pha's peeling behavior might relate to her life history. When she arrived in 1987 in Berlin she was still partially bottle-fed by elephant keepers. Her main caretaker (R.P.) consistently fed her with peeled bananas, and peeled directly in front of her. Pha was never formally conditioned to peel bananas and we suggest that she acquired peeling by observational learning from her human caretakers.
Elephants have outstanding trunk control and even make tools with their trunks 8 . They have been shown to prepare tools to scratch themselves, repel insects, or block roads 8 . Our observations on Pha's banana peeling fi t into this picture. Banana peeling by Pha (22 ± 12 s SD) was faster than humans (89 ± 19 s SD) ( Figure 1H; Mann-Whitney-U test p < 0.001). Hence, Pha's haptic strategy is remarkably effi cient, as humans are highly skilled manipulators. This impressive performance fi ts with the rich neuronal innervation of the trunk 9 and its complex motor control 1,2 . Behavioral studies have also documented a high tactile sensitivity of the elephant trunk 10 . A collection of properties (individuality, dexterity, speed, ripeness specifi city, social modifi cation and the putatively human origin) rather than a single behavioral element make the peeling behavior so distinct. Presumably human-derived complex manipulative behaviors like this were unknown in elephants or, to our knowledge, most other species. Our case study on Pang Pha's banana peeling provides insights about elephant manipulation and cognition not available from other, conventional routes of investigation.

SUPPLEMENTAL INFORMATION
Supplemental information includes supplemental experimental procedures, supplemental references and one video and can be found with this article online at https://doi.org/10.1016/j.cub.2023.02.076.