When 12 men gathered at the Green Bank Observatory in West Virginia to discuss the art and science of alien hunting in 1961, the Order of the Dolphin was born. A number of the brightest minds from a range of scientific disciplines, including three Nobel laureates, a young Carl Sagan, and an eccentric neuroscientist named John Lilly—who was best known for trying to talk to dolphins—were in attendance.
It was Lilly’s research that inspired the group’s name: If humans couldn’t even communicate with animals that shared most of our evolutionary history, he believed, they were a bit daft to think they could recognize signals from a distant planet. With that in mind, the Order of the Dolphin set out to determine what our ocean-going compatriots here on Earth might be able to teach us about talking to extraterrestrials.
Lilly’s work on interspecies communication has since gone in and out of vogue several times within the SETI (Search for Extraterrestrial Intelligence) community. Today, it’s back in fashion, thanks to new applications of information theory and to technological advancements, such as the Cetacean Hearing and Telemetry (CHAT) device, a submersible computer interface that establishes basic communication with dolphins. The return to dolphins as a model for alien intelligence came in 1999, when SETI Institute astronomer Laurance Doyle proposed using information theory to analyze animal communication systems, particularly the whistle repertoire of bottlenose dolphins.
Since Lilly’s initial experiments, researchers have found that a number of species communicate using something that approaches the complexity of human language. Whether it is appropriate to characterize animal communication systems as “languages,” in the same way that English or Mandarin are languages, is a matter of debate. The crux of the debate centers on defining just what constitutes human language.
For one, languages are not innate but acquired through culture. And according to linguists, generally most, if not all, natural human languages enable individuals to refer to abstract concepts or things not present in the immediate environment, to create new words, and to create an infinite number of grammatical sentences of infinite length. Most researchers believe that dolphin squeaks and whistles lack many of these linguistic characteristics. Nevertheless, Doyle argued, their communication is still useful as a model for alien communication. Bottlenose dolphins, for instance, use something called referential signaling, which means that certain communications signals (auditory, visual, or otherwise) correspond to particular aspects of their environment. Some argue that dolphin signals can even be used to convey things such as mood, sex, or age of the dolphin. Their utterances—or squeaks and whistles—may not be as linguistically intricate as ours, but they can convey abstract information.
Doyle confirmed that dolphin signals weren’t random noise by turning to the work of Harvard linguist George Zipf who, in the 1930s, had found a striking pattern common to human languages: The most frequently used word in most languages occurs twice as often as the second most frequently used word, three times as often as the third most frequently used word, four times as often as the fourth most frequently used word, and so on. In American English, for example, the most commonly used word is “the” and the next most frequent word is “of,” which account for about 7 and 3.5 percent of all word usage, respectively.
Satisfying Zipf’s law, says Doyle, appears to be necessary for complex communications, but it’s not sufficient.
When these words are plotted logarithmically on a graph, the relationship among word frequencies yields a line with a slope of -1. Zipf found that the -1 slope is common among most written and spoken languages, from Spanish to Mandarin—a relationship now known as Zipf’s law. Such a formula allows researchers to differentiate meaningful signals from random noise. If a series of sounds had no semantics, its distribution plot would be a flat line, or a slope of 0, because each “word” would be equally likely to occur. A slope steeper than -1, on the other hand, indicates a level of redundancy too high for a human language. Satisfying Zipf’s law, says Doyle, appears to be necessary for complex communications, but it’s not sufficient.
Earlier research had shown that dolphins use a wide variety of signals, but the scientists were unable to determine whether they qualified as something near equivalent to human language. If they did, then their signals should at least conform to Zipf’s law. To test this, Doyle and some of his SETI colleagues looked at signaling from a handful of different species, ranging from squirrel monkeys to cotton plants, as well as dolphins. The tricky part was figuring out how to break down each species’ signals into analyzable units. For dolphins, the researchers looked for natural breaks: spaces between the squeaks and whistles in which there was no sound. Then they checked their frequencies against Zipf’s law.
If the dolphins engaged in meaningful communication with near-human complexity, the frequency of these sounds would yield a logarithmic slope of -1—just like most human languages. So Doyle and his colleagues plotted recordings from a group of captive bottlenose dolphins that had been observed from infancy to adulthood. The resulting slope had a gradient of -.95. This suggests that “dolphinese” may exhibit syntax, says Doyle. “Why would such syntax exist? For one thing, this syntax enables the recovery of errors in the transmission, which definitely has survival value,” he says. “A human example might be the recovery of missing letters in a poorly copied manuscript by the use of spelling rules.”
In comparison, the squirrel monkeys’ Zipf slope was never lower than -0.6, meaning the signals were too random to exhibit syntax. The cotton plant, which communicates through chemical emissions, had a signal distribution slope closer to -1.6, meaning the signals were too redundant.
What Doyle and his colleagues demonstrated is that communication exists on a complexity spectrum. This mathematical tool could be the first step toward a SETI intelligence filter, helping astronomers determine whether intercepted cosmic noise bears linguistic hallmarks. As work by Doyle and his colleagues indicates, perhaps the best place to start is in the watery alien worlds of our own planet. Otherwise, we risk dismissing the first interstellar “hello” as meaningless noise.
Daniel Oberhaus, a freelance writer based in New Delhi, has written for VICE, Slate, PopularMechanics, and The Atlantic. Follow him on Twitter @DMOberhaus.