Koshik is a 12-year-old male Asian elephant housed at the Everland Zoo in South Korea.  For some years he had been a local star that was the subject of some internet fame due to his uncanny ability to produce human-like sounds.  Not only this, but they actually seemed to resemble Korean words.

In a new paper in the journal Current Biology by Angela Stoeger and colleagues, Koshik’s vocalizations were put to the test.  Could he really produce words, as trainers claimed?

The researchers recorded Koshik’s special utterances and played them back to a panel of native Korean speakers who had never heard him before. These participants did not know who was producing the sounds or what they were supposed to mean. They were then asked to write down the words they heard.  They found that Koshik’s call resembled five Korean words: ‘‘annyong’’ (hello), ‘‘anja’’ (sit down), ‘‘aniya’’ (no), ‘‘nuo’’ (lie down), and ‘‘choah’’ (good).  He appeared to be very good at reproducing the vowels in each of the words, but the consonants were more problematic.  “Choah” for instance was interpreted sometimes as “boah” (look) and “moa” (collect) by the human listeners.

A comparison of Koshik’s vocalizations and that of his trainer as well as another person. The dark regions are known as spectral peaks, or formants in the case of the human words.

Next the researchers took apart the acoustic features of the vocalizations themselves, comparing them to utterances by his trainer and another unassociated human male as well as hundreds of natural vocalizations produced by elephants in other zoos and in the wild.  Quite clearly the measurable features of Koshik’s word-like sounds more closely resembled the human speech than the natural calls of elephants. More technically, he was quite good at matching the pitch and the first and second ‘spectral peaks’ to the human pitch and speech ‘formants’. Pitch can roughly be thought of the frequency or tone of a sound.  Elephants naturally produce calls that are either very high-frequency or very low-frequency (sometimes infrasonic, or below the range of human hearing), due to the structure of their vocal tract and sound chambers.  Spectral peaks and formants are regions where the sound is shaped by the mouth and (usually) tongue to emphasize certain parts over others. These primarily account for the differences between vowels such as “o” and “a”.  This is quite a feat for an elephant, since clearly the vocal apparatus of an elephant is quite different to that of humans!  Koshik seems to have overcome this difficulty by learning to place his trunk inside his mouth and somehow manipulate the inside manually (for lack of a better word!). Just imagine that in order to speak a foreign language you had to stick your fingers in your mouth and muck about with your insides to produce the appropriate sounds – now you have some idea of what a tricky thing this is to do!

What prompted Koshik to do this? It appears that he was never especially trained to produce these remarkable sounds, he just started doing them spontaneously. It may have taken him a few years of practice before his trainers even recognize what he was doing (a video released some time ago shows him practicing all alone in his shelter while being recorded by a remote camera).  Researchers speculate that perhaps the fact that he was the only elephant at Everland from the age of five onwards had something to do with it. They hypothesize that perhaps these vocalizations were an attempt to make social bonds for an animal that had been isolated from its own species.

Two males resting under a tree in Uda Walawe national park, Sri Lanka.

This is an interesting hypothesis, considering that Koshik is a male. In the wild, females tend to be found in social groups, but young males of his age leave their mothers and start to roam further afield. While males can often be spotted alone, they do also share the company of other males and sometimes even can be found keeping amiable company with females and herds. This suggests that males too may seek to social companionship. There are other possibilities of course – perhaps this was merely a novel and challenging way of playing for Koshik, and nothing to do with whether or not he had company.

Why is it so noteworthy that Koshik was able to make speech sounds in the first place? First, vocal mimicry is relatively rare in the animal kingdom. Birds are the most common example, and bird song is thoroughly studied because of the insights it can provide into how vocal learning and copying occurs. But it is extremely rare among mammals. Numerous studies of primates, for instance, have failed to show that the ability to produce and modify sounds can be learned, although they can learn to apply naturally produced calls to specific situations. Only a few instances have been documented in other species. One previous case involves a beluga whale who could apparently say his name (‘Lagosi’).  Coincidentally, in a previous issue of the same journal, was an article on another beluga who produced human-like sounds.

Does Koshik know what he is saying? In other words, do his utterances have any meaning for him? Probably not – he was trained to produce certain behaviors in response to the commands uttered by his keepers, but it is not likely he expects his keepers to reciprocate when he utters them instead! Although, if they DID so, he might learn that it’s a possibility – even children have to learn how to use words.  Another intriguing question is whether elephants in the wild naturally have to learn any part of their vocal repertoire. The ability to learn how to produce certain signals (in this case, acoustic signals) and then to apply them under appropriate circumstances is the rudimentary basis of language. While this study is far from showing that elephants have any linguistic prowess, it remains a fascinating area to investigate.

Stoeger, A., Mietchen, D., Oh, S., de Silva, S., Herbst, C., Kwon, S., & Fitch, W. (2012). An Asian Elephant Imitates Human Speech Current Biology DOI: 10.1016/j.cub.2012.09.022