Talking With Your Mouth Full: The Feeding Calls of the Humpback Whale

The Native American tribe known as the Tlinket of southeast Alaska have many stories about how raven stole the Sun, how bear and ant argued about the constellations, and how the wolf and moon are related. But as far as I know, they don't have any stories about the humpback whale. If they did, they might have told how the humpback gave voices to all the other animals, for this creature has all the voices of the animal kingdom.

We have recorded humpbacks making sounds like the trumpeting of elephants, roars like lions, whistles like dolphins, clicks like the sperm whale, mooing like cows, chattering like monkeys, and several very human-like vocalizations--some even sounding like an unusual language, with exclamations like "whoops!" Although we are just beginning to document and classify all the diverse sounds of the humpback whales, we already expect its repertoire to exceed that of any other animal we have studied to date.

The humpbacks of southeast Alaska are known to migrate thousands of miles to Hawaii to mate, where they sing long and complex songs. Several variations of these songs are started at the beginning of the season, then eventually all humpbacks are singing the same song for that year. They head to Alaska in the summer to feed, and here they have a different kind of vocalization based more on feeding needs. However a colleague, Chris Gabriele of Glacier Bay National Park System, has also found that humpback whales still sing even in this feeding season.

Some anthropologists have suggested that the increasing complexity of human society might have led to the complexity of human language. One might test this for a simpler case, one independent of written systems, etc., by comparing animal communications. If the same, or closely related species, might be compared under different conditions of complex versus more isolated social circumstances, one could see if there is, indeed, such a correlation of these social conditions with communication complexity.

The humpback whales of Chathum Strait and Fredrick Sound could provide such a test. The former feed by making bubble nets around herring, while the humpback whales of Fredrick Sound eat krill. Krill are pretty easy to catch--the whales just more-or-less guide the krill into their baleen-filled mouth with a large fin. However, building a complex, cylindrical net out of bubbles to capture a school of herring is another matter altogether.

The Chathum Strait humpback whales--up to 30 or more--will group together to produce such a bubble net. Our collaborators, Fred Sharp and Pieter Folkens of the Alaska Whale Foundation, have discovered that the humpbacks that make the bubble nets are not actually directly related. In other words, based on DNA testing (of the skin they shed, picked up off the water), the offspring of the males and females who fish together by making a coordinated bubble net are not actually parts of the same family group.

It is known that matriarchal elephants teach their offspring the places to forage. As another example, families of wolves teach offspring how to hunt. But no other animal to date, besides humans, has been known to form a lasting social relationship based on a (sort-of) profession, among non-family members. Perhaps the humpback whales may be closer socially to humans than any other animal.

But how does one actually measure the complexity of humpback vocalizations? With colleagues Brenda McCowan and Sean Hanser of the University of California, Davis, we have been applying the mathematics used for calculating the amount of information sent through computer lines to quantify the amount of information vocalized between humpback whales (and other species). This field, incidentally, is known as "information theory," and was developed by Claude Shannon of Bell Labs several decades ago. Information theory measures not so much what is being said--that is, it does not lead directly to the meaning of the signals. Rather, it precisely measures what a particular communication system could be saying--in other words, what that system of communication's "carrying" capacity could be--what it is capable of saying. We have found that some animal communication systems are simpler than others--dolphin whistles are apparently more complex than ground squirrel alarm calls, for example.

The information theory measure that can best be used to quantify complexity is known as the "entropy." It is a measure of the number of choices available in a given communication system. Suppose we learned the communication system of squirrel monkey chuck calls. Might we ever be able to translate Shakespeare into squirrel monkey? By comparing the entropy of the two samples of communication, we would be able to conclude whether this might even be possible. From our studies to date, we would conclude that this would not be possible at the current complexity of the chuck call vocal communication system compared to the complexity of the human vocal communication system.

An important measure of entropy is the highest "entropic-order" at which the communication systems peaks. In measuring this, we ask how dependent the signals are on each other. In human speech we have grammar and in human writing we have spelling (or brush strokes, etc.) that depend on each other. If you made a copy of a written page, but the toner in the copy machine was low, you would find that you could nevertheless recover some of the missing words because there are rules of spelling and grammar superimposed on our language system. It is these rules that allow error recovery - and this works in both vocalization as well as written communication systems (as well as any others, e.g., chemical signaling units, bee dances, visual facial features, etc.)

Therefore, we have used information theory to measure the vocal communications between feeding humpback whales, including the degree of complexity (or "syntax" if you like). We have also measured feeding calls when boats were in the area, and when there were no boat engine noises in the background. It turns out that the humpback whales increase the dependence of their signals on each other (decreasing the rate of information transfer, but increasing the degree of error recovery possible) when boat noise interferes with their feeding calls being heard. On average, the boat noise in Glacier Bay, for example, would decrease the amount of information that could be transmitted through the water by about one-quarter, and this was how much lower the humpback whale vocalization rate of information was also measured to decrease. In other words, the humpbacks seemed to be responding to the boat noise (a quantification that will allow boat noise to be regulated better in Glacier Bay). But the humpbacks also seemed to know just how much to compensate--on average they did not over or under-correct for the boat noise either.

We hope to be able to quantify the vocal feeding behavior of the humpback whales in Chathum Strait as compared to those in Fredrick Sound.

And how might this apply to the search for extraterrestrial intelligence? If there is a relationship between social complexity and vocal complexity, then the measure of one will be a measurement, to some degree, of the other. If a SETI signal is received, and is a normal (i.e., un-coded) communication, it will have to obey the rules of information theory in order to transmit information. Thus, a measure of the information complexity of such SETI signals could also be a first direct measurement of the social complexity of an extraterrestrial species, irrespective of the actual decipherment of the meaning of such a message itself. Exciting prospect indeed!