Elephants Say "Bee-ware!"

What information is contained in the call of a mammal? Some calls might reflect the internal emotional state of the animal, like fear or anxiety, or they can refer to an external object, agent, or event, like the presence of a predator. Rhesus monkeys, lemurs, baboons, and guinea pigs, for example, will produce calls when separated from their conspecifics or in the presence of a stranger. Howler monkeys produce specific alarm calls for avian predators, even when they have never encountered an avian predator for several generations. Vervet monkeys produce different calls in response to leopards and eagles, suggesting that the acoustic signals carry information like "predator from the ground" or "predator from the sky." The yellow-bellied marmot's alarm calls are not predator specific, but increase in rate along with the level of perceived risk. Acoustically, typical features of alarm calls include changes in tempo (rate and duration), source feature (frequency and amplitude), or filter features (vocal tract modulations).

Unlike other mammals, African elephants have few predators that pose significant risk to their survival. In Kenya's Amboseli National Park, elephants have been observed retreating and posturing defensively in the presence of Masaai tribesmen, who have been known to kill elephants. However, they display the same response to auditory playback of unfamiliar conspecifics, suggesting that this behavioral response may be a more general response to strangers instead of a specific response to threat.

Figure 2: The starring characters in this study. Dumbo the flying elephant, and Spike the bee.

More recently, it has been observed that African elephants avoid contact with African honeybees, even avoiding feeding on trees which contained beehives. A recent study conducted by Dr. Lucy King, and colleagues (and partially funded by Disney!) was designed to further investigate this initial observation.

First, they recorded the sounds of angry bees, and played them to groups of elephants. Some elephants were played white noise instead. While they did this, they recorded the vocal responses of those elephants. Would their vocalizations in response to the bees be different than in response to white noise?

So how did the elephants do?

Elephants moved away in response to the bee sounds, farther than in response to white noise, and faster (panels A-B). Also, during the bee playback, the elephants displayed increased rates of headshaking and dusting (kicking up dust with their feet), compared to white noise. The total number of calls was also significantly greater in response to bees than to white noise. Family size, age, family composition, temperature, time of day, altitude, and air pressure were all counted out as possible confounding variables. We'll come back to panels C-D.

Next, the researchers measured the acoustic properties of all vocalizations produced in response to the bees and to white noise. Rumbles produced in response to white noise and bee sounds had increased fundamental frequency and fundamental frequency range, compared with control rumbles (pre-recorded, spontaneously produced vocalizations). Also, rumbles produced in the bee trials showed a shift in the location of the second formant, when compared with both the white noise and control rumbles, suggesting that specific bee-related information is contained in the location of the second formant. Potential confounding variables, such as body size or physical exertion, were ruled out.

How to test this? Does the location of the second formant contain critical information?

They took the recordings they made in the first experiment - of elephants responding to bee sounds or to white noise - and played them back (along with the control recordings) to other elephant groups. Was there something in the vocalization produced in response to the bees that would provide specific information to other elephants about the apparent threat? Are the vocalizations made by elephants in response to bee sounds threat-general or threat-specific?

They played three types of recordings back to seventeen elephant families. The bee rumble was simply the recording made by the elephants in response to the playback of angry bees. The control rumbles were other generic rumbles made by the same elephants, matched for duration and amplitude to the bee rumble. The white noise rumble was actually the bee rumble, acoustically modified so that the second formants were lowered in frequency location to resemble rumbles produced in response to white noise playbacks. Formant location shifts are produced due to modulations of the vocal tract.

On the left is the elephant vocalization in response to the bee sounds. On the right, the second formant is artifically lowered to match the rumbles produced in response to white noise, to produce the "white noise rumble".

By comparing the elephants' responses to the bee and control rumbles, the experimenters could determine if there was bee-specific information contained in the duration or amplitude of the call. By comparing the bee and white noise rumbles, the experimenters could see if there was bee-specific information contained in the location of the second formant (and thus produced by specific modulation of the vocal tract).

With permission from the authors, I've combined the rumble recordings from the supplementary materials into a video clip. Each type of rumble repeats twice:

And here is the response of one elephant family to the playback of the bee rumble:

They moved farther away and faster in response to the bee rumble than either the control or white noise rumbles (panels C-D, Figure 3, above). There is, however, a somewhat muted effect when comparing the responses to the actual bee sounds and the bee rumbles. As when responding to the playback of the actual bee sounds, the elephants displayed increased headshaking, but not increased dusting.

Overall, the behavioral responses of the elephants are consistent with defensive posturing against bees. Headshaking and dusting would knock bees away, and fleeing from the area would reduce the risk of being stung. Despite the fact that elephants have thick skin, they are particularly sensitive to bee stings around the eyes and trunk. Also, juvenile elephants have not yet developed the thick skin characteristic of adult elephants.

The rumbles produced in response to the bees may reflect an emotional state (e.g. fear), they could serve to coordinate group movement (social communication, potentially threat-specific), or they could function as alarm calls to more distant elephants (social communication, potentially threat-general). They could also be used for social facilitation - to teach the younger more inexperienced elephants about a relevant threat.

The acoustic changes in response to bees - increased and more variable fundamental frequencies - are consistent with data from the alarm calls of other mammals more generally. But the bee response was also characterized by the upward shift in the location of the second formant, which isn't generally associated with emotional arousal. The authors point out that, in humans, modulations of the physical properties of the vocal tract result in the production of different vowel sounds, which are associated with different semantic (meaning-related) information. If elephant vocalizations are functionally similar, then the formant shift found in response to bees may also carry semantic, and thus threat-specific, information. Since the elephants who were played the bee rumbles displayed the defensive posturing behaviors consistent with the actual presence of bees, this provides converging evidence for this threat-related alarm call. However, dusting behavior was seen during the bee sound experiment, but not the bee rumble playback experiment. It therefore not clear if the bee rumble is threat-general or specific. More research is needed to tease these possibilities apart.

Why is this research important? Aside from the general importance of better understanding animal communication, as it can inform our understanding of human communication, this research has very practical implications as well. Elephants regularly raid the crops of humans; strategically placed beehives (or even just speakers broadcasting bee sounds or bee rumbles), could minimize human-elephant conflict and potential elephant deaths.

Want to read more about this research? http://www.elephantsandbees.com/
And thanks to Disney for partially funding this cool research!

King, L., Soltis, J., Douglas-Hamilton, I., Savage, A., & Vollrath, F. (2010). Bee Threat Elicits Alarm Call in African Elephants PLoS ONE, 5 (4) DOI: 10.1371/journal.pone.0010346