Among nonhuman animals, birds take first place in imitating sounds, including human words. But perhaps surprisingly, marine mammals place a respectable second. Certain harbor seals have learned to bark out (barely) recognizable words, and whales imitate each other during the development of long, complex songs. But the most widely studied marine mammal vocalizers are dolphins.

Dolphins communicate with a great number of different sounds, from the repetitive clicks used for echolocation (and possibly to communicate) to whistles and grunts; in captivity, they can imitate human words to some extent. Their echolocation appears truly remarkable. A blindfolded dolphin can find an object the size of a penny on the bottom of a swimming pool and can distinguish small objects based on their shape and the material they’re made of. But more remarkable, perhaps, are the vocalizations more obviously used for communication.

Bottlenose dolphins (Tursiops truncatus) produce signature calls.

Photo © 2006 Vincent Janik

Whistles at first appear to say one simple thing: “Hi! I’m Flipper. Hi! I’m Flipper.” Research since the 1960s has indicated that dolphins each have a unique whistle, called a signature whistle. This implies dolphins can produce several different whistles—at least 10 to 25, the number of individuals in an average-sized group. Furthermore, they must learn the signature whistles of every other dolphin in the group. Reviews of the whistles of more than 100 dolphins suggest they do not choose a signature from a fixed set of whistles, but develop an individual whistle. As the dolphin matures, its signature becomes somewhat stereotyped. To identify individuals by their whistles, dolphins must also have the mental ability to link the individual whistle with an individual dolphin. This seems likely, judging from experiments teaching dolphins to link human hand signals, human whistles and even electronically generated sounds to particular objects (for example, imitating the tune to “Mary Had a Little Lamb” when shown a Frisbee).

Behaviorists Peter Tyack of Stanford University and Laela Sayigh of the University of North Carolina at Wilmington studied a population of wild dolphins near Sarasota, Florida. These dolphins, a group studied for decades by Randal Wells of the Chicago Zoological Society, remain in one area, so researchers can repeatedly find individuals, which they recognize by the dolphins’ individual markings. Because they can’t tell which dolphin is making a particular sound in a group, they record the dolphins individually after corralling them with a net. The technique provides only a seminatural environment, but has yielded intriguing information.

First, Sayigh confirmed that individual dolphins have recognizable signature whistles, and that those whistles remain stable for at least a decade. She also found that mothers and calves remain in vocal contact when one is in the corral and the other swimming nearby. When Sayigh began recording newborn calf whistles, she found a faint, quavery sound that varied from whistle to whistle, like the written signature of a young child. By the age of one, the calf had firmed up an individual whistle that then remained more or less constant for a decade.

Sayigh also found male and female calves develop whistles differently. A female calf learns a whistle distinctly different from her mother’s. A brother’s whistle, on the other hand, develops from a variable baby whistle to an adult form closely resembling his mother’s. In this population of dolphins, at least, females with newborn calves hang out with their mothers, grandmothers and other females, forming a long-lasting group. Similar signature whistles among a young female, her mother, grandmother and even other females could lead to confusion, like having a family with two members with the same name; the wrong person is always taking the telephone call. Male calves leave the group when they mature, so they stand little chance of being misidentified.

While this modification of signature calls appears to rely on learning, Sayigh has not ruled out a genetic basis. For example, if the signature call were genetic, it might reside on in the X chromosome. A female would inherit two X chromosomes, one from her father and one from her mother. Her signature would then differ from either. A male, on the other hand, inherits only one X, from his mother, and his call would resemble hers.

Studying signature whistles can pose problems for researchers. Although humans can hear dolphin whistles under water, they cannot locate the source of the whistles. Nor can they locate the source of the whistles picked up by underwater microphones and broadcast topside. Isolating the dolphins solves the problem of identifying who’s squeaking. But even when two dolphins can communicate by way of a sort of underwater walkie-talkie—as they could in one experiment—isolation severely disrupts social communication.

Peter Tyack invented one solution to the problem, allowing him to reliably record the whistles of two or more individual dolphins. He calls his invention the vocalight. The three-and-a-half-inch-long device looks like a cross between a small flashlight and a model of Star Trek’s U.S.S. Enterprise. A suction cup holds the device to the slick dolphin skin in front of the blowhole. A row of light-emitting diodes points to the front and a battery case to the back. Inside, a microphone picks up the dolphin’s sounds and electronics illuminate the diodes in response to the sounds. The louder the sounds, the more diodes light up.

Tyack first studied two dolphins, Scotty and Spray, who lived in an aquarium called Sealand in Brewster, Massachusetts. Tyack set up the two with vocalights, one with red diodes and one with green. He and other observers called out the color and number of diodes whenever they heard a whistle picked up by an underwater microphone. A tape recorder captured both the whistles and the observers’ voices.

By displaying the whistles as sonograms, Tyack found more than three quarters fell into two easy-to-recognize categories. The Type 1 whistle rises, swoops down to a lower pitched half-second-long whistle, then rises slightly. The Type 2 whistle also rises, falls and levels out, but then rises abruptly at the end (see figure XX). Although both dolphins produced both kinds of whistles, Tyack noticed individual variations. Spray’s Type 1 was not identical to Scotty’s Type 1, and their Type 2 whistles differed even more. Interestingly, the two dolphins split their whistles between the two types differently. Two thirds of Spray’s whistles fell into the Type 1 category, while almost three quarters of Scotty’s were Type 2. Tyack interprets these results to mean that Type 1 was Spray’s signature and Type 2 Scotty’s. Each dolphin imitated the other’s signature whistle some of the time, Tyack contends, perhaps using the other dolphin’s signature whistle as a label or name.

The dolphins also whistled variations on those two themes—they left parts out, varied the duration of the whistle, changed the pitch slightly and varied the shape of the sounds as seen on a sonogram. Tyack maintains the dolphins could detect these variations and the many whistles that fell into neither category.

Spray unfortunately died, leaving Scotty in isolation. Tyack returned after two years, curious to see if Scotty’s vocalizations had changed. Scotty no longer used Spray’s favorite whistle at all. And his whistles in general had become quieter and briefer, lending weight to the idea that dolphin whistles function as social communication.

Given this evident complexity, dolphins’ vocabularies may well encompass much more than “Hi! I’m Scotty.” and “Hi! I’m Spray.”

Although dolphin vocalizations in the wild may stretch beyond mere signature whistles, past attempts to read imitation of human sounds into their squeaky voices stand on shaky ground (see sidebar: John Lilly). Even today, rapid evaluation of their high-pitched sounds proves difficult. Furthermore, their sleek, fingerless flippers make it difficult for them to produce the kinds of artificial communication—hand gestures and keyboard use—that apes can muster (see chapter 12). But research has moved ahead on the other side of communication, dolphins’ ability to understand language.

Louis M. Herman and his colleagues at the Kewalo Basin Marine Mammal Laboratory, University of Hawaii at Manoa, have taught four wild-caught dolphins artificial languages. One language consists of computer-generated, high pitched “words.” In the second, a trainer conveys words with hand and arm gestures. Neither language bears any resemblance to human language except in that both contain grammatical rules. Each language comprises about 40 words, nouns such as “channel, gate, person” and “ball;” verbs such as “under” (the go implied) and “fetch;” and modifiers such as “surface, bottom, right” and “left.”

Using these languages, Herman can ask a number of questions about dolphins’ ability to comprehend words as referents to specific objects and to understand that the order of the words makes a difference. Unlike human grade schoolers, the dolphins never received any grammar lessons. Instead they learned the rules by example. “Phoenix Ake under,” for example, means the trainer wants Phoenix to swim under Akeakamai (Ake for short). The gestural language learned by Ake included a reverse grammar, to prevent word-by-word responses. The dolphin receiving a gestural command must understand the full two- or three-word sentence before beginning to respond.

Herman found that the dolphins learned both what each word referred to and how to interpret the order of the words. For example, Ake distinguished between gestural sentences such as “right hoop left Frisbee fetch” and “left hoop right Frisbee fetch.” The first sentence means “Take the Frisbee on your left to the hoop on your right.” Success with such sentences demonstrates the dolphins’ ability to understand both the semantic (word meaning) and syntactic (sentence pattern) components of the language, Herman writes.

Ake, however, has gone beyond merely responding as trained. She even invents her own logical responses to unusual situations. If a trainer commands “Frisbee hoop in,” a two-gesture sequence asking Ake to put the hoop on top of the Frisbee, Ake normally complies. But Ake has also learned to use yes and no paddles. For example if either the Frisbee or the hoop were unavailable, the trainers would expect her to press the no paddle. Sometimes, asked the above question with both objects in the tank, Ake would put the hoop—the object the command asks her to manipulate—on the yes paddle, a behavior she invented. If the Frisbee is not in the tank, Ake will put the hoop on the no paddle. If, on the other hand, the hoop is missing, she will not move the Frisbee—the trainer didn’t ask her to—but will press the no paddle, meaning she can’t comply with the command.

Ake responds to novel combinations of words correctly, showing understanding of both the words and the sequence. But what does she do with a construction that makes no sense? Ake will move an object to make it possible to comply with a command. She will lift a hoop that’s lying on the bottom of the tank so she can swim through it, for example. If instead she sees a sentence that gives an impossible command, such as “person water fetch,” Ake does nothing. She can’t move the water to the person. A more complex kind of “impossible” command, “person water hoop fetch,” contains too many nouns. In this instance Ake sensibly ignores the noun water and interprets the sentence as “take the hoop to the person.” She has interpreted the impossible “water” as a mistake, much like a human will assume a duplicated duplicated word represents a typographical error. Herman sees these untaught responses as indications that dolphins make a mental representation of the grammatical rules of their artificial language. By referring to this memory of the structure, they can make sense of novel or even nonsensical sentences.

A picture is worth a thousand words, the saying goes. In this digital age, is a picture composed on only every other pixel worth only five hundred? On the contrary, we can still “read” a picture with a large part of the digital detail missing, an image of Abraham Lincoln made of only a few dozen blocks of color, for example. Herman wondered whether Phoenix and Ake could “read” gestural communication with reduced detail. First he presented a videotape of a trainer to the dolphins on an underwater TV monitor. They responded almost as well as to the real trainer. Next he blanked out the head and torso, leaving only the arms, then removed the arms, and finally presented only two moving white spots to represent the hands. The dolphins responded to all of these presentations, their correct responses falling off only with the last, most abstract presentation. Even this minimalist communication elicited more correct responses than chance would predict. Human trainers (college students) with four months of experience in the gestural language did about as well.

From these and other studies, Herman concludes that dolphins use words of the artificial language to refer to objects in an abstract way, and can make sense of the artificial grammar as well.