Resource Guide Links:

language learning

dolphin perception

the bottlenose dolphin: natural history and ecology

getting to know humpback whales


tdi for kids



Language Learning

Louis M. Herman
Kewalo Basin Marine Mammal Laboratory, University of Hawaii
and The Dolphin Institute, Honolulu, Hawaii

Human language and ape language

No single trait has been linked more closely with the human species than has language. However, the definition of language and its uniqueness as a human trait continue to be areas of study and debate. Some, such as the linguist Noam Chomsky, take an evolutionary discontinuity position professing that language is a highly nunique adaptation supported by special modifications of the brain that appear only in humans. Others, such as the anthropologist Barbara King, favor a continuity position, which suggests that language must have its roots in earlier hominoid adaptations for
communication, and that some of these adaptations may still be extant in modern ape species. The work on teaching various language-like systems to apes by Beatrice and Alan Gardner, David Premack, Duane Rumbaugh and others, beginning in the mid-1960s and continuing throughout the decade of the 70s, seemed to provide a genuine link between human and ape in fundamental language competency. This early work reported that chimpanzees were able to learn to use and understand not only words but also sentences. Sentences give human language its great communicative power, through the infinite variety of meanings that can be constructed by the recombination of words. To understand a sentence the human listener must take account of both the meaning of the words and their grammatical relationships to one another, as governed by word order or other syntactic devices. This early work on teaching language to apes was thrown into disarray, however, by additional studies and criticisms from other researchers, such as Herbert Terrace and Carolyn Ristau. These researchers argued that the putative “sentences” produced by the apes were largely an artifact of context, imitation, or cueing. In particular, although sequences of words were indeed produced by the apes, the sequences had no syntactic structure that enhanced, explained, or modified meaning.

Until recently, the work with apes was focused on language production and paid scant attention to language comprehension. Investigators attempted to teach the apes to produce language—where words were represented by gestures, keyboard symbols, or other types of artificial symbols. These investigators assumed that if the ape produced a word, or a series of words, that it therefore understood what the word or sequence represented. They also assumed that the ape could understand those same words or sequences when produced by the human partner. These assumptions, when finally tested, proved false. It was shown, instead, that comprehension did not automatically flow from language production. The preeminence of comprehension in language development, only recently appreciated in the ape language field, has long been emphasized among those studying child language. Language comprehension by young children develops earlier than language production, and even into adulthood comprehension vocabularies exceed speaking vocabularies.

Recent work with bonobo chimpanzees, pioneered by Sue Savage-Rumbaugh, has emphasized language comprehension and has progressed well beyond the earlier ape language studies. The bonobos have shown an ability to learn to understand instructions given in spoken English sentences. Together with some of her earlier work, Savage-Rumbaugh has shown that chimpanzees can learn to appreciate the symbols (words) of the language “referentially.” The understanding that words refer to things or events in the real world is one of the key characteristics of human language. Among other things, referential understanding enables us to discuss things that are not immediately present or that happened at a different place or time.

Dolphins and Language

Natural language? Dolphins produce various types of sounds, including clicks, burst-pulse emissions, and whistles. Clicks are used for echolocation, the dolphin’s form of sonar. Through echolocation, the dolphin can examine its world through sound, by listening to the echoes returning from objects struck by the clicks. Burst-pulse sounds may indicate the dolphin’s emotional state, ranging from pleasure to anger. However, these type of vocalizations have been little studied and much remains to be learned about them. Whistles may be used for communication, but it is still an open question as to whether, or how much, of whistle communication is intentional versus unintentional (e.g., rapidly repeated whistling may be elicited by stress, without any specific intention to convey that emotional state to others). During the 1960s, researchers attempted to determine whether the whistle vocalizations might be a form of language. Investigators recorded whistles from many dolphins in many different situations, but failed to demonstrate
sufficient complexity in the vocalizations to support anything approaching a human language system. Some of the early work instead pointed to the stereotypy of the whistles from individual dolphins, leading David and Melba Caldwell to suggest that the whistle functioned principally as a “signature,” with each individual dolphin producing a unique signature. Presumably, this enabled that individual to be identified by others. Other researchers have noted, however, that there can be a great deal of flexibility in the whistle. Douglas Richards, James Wolz, and Louis Herman, at the Kewalo Basin Marine Mammal Laboratory at the University of Hawaii, reported a study showing that a dolphin could use its whistle mode to imitate a variety of sounds generated by a computer and broadcast underwater into the dolphin’s habitat. Peter Tyack later reported that one dolphin could imitate another’s whistle, thereby possibly referring to or calling that individual. As was noted earlier, referring symbolically to another individual, or to some other object or event in the environment, is one of the basic characteristics of a language. However, we still do not know to what extent the dolphin’s whistles may be used to refer to things other than themselves or another dolphin. This is a fruitful area for additional study, however.

Although the evidence strongly suggests that dolphins do not possess a natural language, like the case for apes, it is still important and informative to study whether dolphins might nevertheless be able to learn some of the fundamental defining characteristics of human language. Any demonstration of language-learning competency by dolphins would bear on questions of the origins of human language, shifting the emphasis from the study of precursors in other hominoid species to common convergent characteristics in ape and dolphin that might lead to advanced communicative and cognitive capacities.

Early attempts at teaching language to dolphins. From the mid-1950s to the mid-1960s, John Lilly promoted the idea that bottlenosed dolphins (Tursiops truncatus) might possess a natural language. He based this supposition on this species’ exceptionally large brain with its richly developed neocortex. He reasoned that the large brain must be a powerful information processor having capabilities for advanced levels of intellectual accomplishment, including the development of a natural language. He set about to uncover the supposed language. Failing in that quest, he then attempted, also without success, to teach human vocal language (English) to dolphins he maintained in his laboratories. Dolphins have a rich vocal repertoire, but not one suited to the production of English phonemes. The procedures used by Lilly and the data he obtained were presented only sketchily, making any detailed analysis of his efforts at teaching language moot.

In the mid-1960s, Duane Batteau developed an automated system that translated spoken Hawaiian-like phonemes into dolphin-like whistle sounds that he projected underwater into a lagoon housing two bottlenosed dolphins. He then attempted to use these sounds as a language for conveying instructions to the dolphins. A major flaw in his approach, however, was that individual sounds were not associated with individual semantic elements, such as objects or actions, but instead functioned as holophrases, (complexes of elements). For example, a particular whistle sound instructed the dolphin to “hit the ball with your pectoral fin.” Another sound instructed the dolphins to “swim though a hoop.” Unlike a natural language, there was no unique sound to refer to hit or ball, or hoop, or pectoral fin, or any other unique semantic element. Hence, there was no way to recombine sounds (semantic elements) to create different instructions, such as “hit the hoop (rather than the ball) with your pectoral fin.” After several years of effort, the dolphins were able to learn to follow reliably the holophrastic instructions conveyed by each of 12 or 13 different sounds. However, because of the noted flaw in the approach to construction of a language, the experiment failed as a valid test of dolphin linguistic capabilities.

Kewalo Basin dolphin language studies. The work on dolphin language competencies by Louis Herman and colleagues at the Kewalo Basin Marine Mammal Laboratory in Honolulu was begun in the mid-1970s and emphasized language comprehension from the start. These researchers, working principally with a bottlenosed dolphin named Akeakamai housed at the
laboratory, constructed a sign language in which words were represented by the gestures of a person’s arms and hands. The words referred to objects in the dolphin’ habitat, to actions that could be taken to those objects, and to relationships that could be constructed between objects. There were also location words, left and right, expressed relative to the dolphin’s location, that were used to refer to a particular one of two objects having the same name, e.g., left hoop vs. right hoop. Syntactic rules, based on word order, governed how sequences of words could be arranged into sentences to extend meaning. The vocabulary of some 30 to 40 words, together with the word-order rules, allowed for many thousands of unique sentences to be constructed. The simplest sentences were instructions to the dolphin to take named actions to named objects. For example, a sequence of two gestures glossed as surfboard over directs the dolphin to leap over the surfboard, and a sequence of three gestures glossed as left Frisbee tail-touch directs the dolphin to touch the Frisbee on her left with her tail. More complex sentences required the dolphin to construct a relationship between two objects, such as taking one named object to another named object or placing one named object in or on another named object. To interpret relational sentences correctly, the dolphin had to take account of both word meaning and word order. For example, a sequence of three gestures glossed as person surfboard fetch tells the dolphin to bring the surfboard to the person (who is in the water), but surfboard person fetch, the same three gestures rearranged, requires that the person be carried to the surfboard. By incorporating left and right into these relational sentences, highly complex instructions could be generated. For example, the sequence of five gestures glossed as left basket right ball in asks the dolphin to place the ball on her right into the basket on her left. In contrast, the rearranged sequence right basket left ball in means the opposite, “put the ball on the left into the basket on the right.” The results published by Louis Herman, Douglas Richards, and James Wolz showed that the dolphin was proficient at interpreting these various types of sentences correctly, as evidenced by her ability to carry out the required instructions, including instructions new to her experience. These were the first published results showing convincingly an animal’s ability to process both semantic and syntactic information in interpreting language-like instructions. Semantics and syntax are considered core attributes of any human language.

Ronald Schusterman and Kathy Krieger tested whether a California sea lion named Rocky might be able to learn to understand sentence forms similar to those understood by the dolphin Akeakamai. Rocky was able to carry out gestural instructions effectively for simpler types of sentences requiring an action to an object. The object was specified by its class membership (e.g., “ball”) and in some cases also by its color (black or white) or size (large or small). In a later study, Schusterman and Robert Gisiner reported that Rocky was able to understand relational sentences requiring that one object be taken to another object. These reports suggested that the sea lion was capable of semantic processing of symbols and, to some degree, of syntactic processing. A shortcoming of the sea lion work, however, was the absence of contrasting terms for relational sentences, such as the distinction between “fetch” (take to) and “in” (place inside of or on top) demonstrated for the dolphin Akeakamai. Additionally, unlike the dolphin, the sea lion’s string of gestures were given discretely, each gesture followed by a pause during which the sea lion looked about to locate specified objects before being given the next gesture in the string. In contrast, gestural strings given to the dolphin Akeakamai were without pause, analogous to the spoken sentence in human language. Further, Rocky did not show significant generalization across objects of the same class (e.g., different balls), but unlike the dolphin seemed to regard a gesture as referring to a particular exemplar of the class rather than to the entire class. Thus, although many of the responses of the sea lion resembled those of the dolphin, the processing strategies of the two seemed different, and the concepts developed by the sea lion appeared to be more limited than those developed by the dolphin.

Akeakamai’s knowledge of the grammar of the language. As a test of Akeakamai’s grammatical knowledge of the language she had been taught, Louis Herman, Stan Kuczaj, and Mark Holder constructed anomalous gestural sentences. These were sentences that violated the syntactic rules of the language or the semantic relations among words. The researchers then studied the dolphin’s spontaneous responses to these sentences. For example, the researchers compared the dolphin’s responses to three similar gestural sequences: person hoop fetch, person speaker fetch, and person speaker hoop fetch. The first sequence is a proper instruction; it violates no semantic or syntactic rule of the learned language. It directs the dolphin to bring the hoop to the person, which the dolphin does easily. The second sequence is a syntactically correct sequence but is a semantic anomaly inasmuch as it directs the dolphin to take the underwater speaker, firmly attached to the tank wall, to the person. The dolphin typically rejects sequences like this, by not initiating any action. The final sequence is a syntactic anomaly in that there is no sequential structure in the grammar of the language that provides for three object names within a sequence. However, embedded in the four-item anomaly are two semantically and syntactically correct three-item sequences, person hoop fetch and speaker hoop fetch. The dolphin in fact typically extracts one of these subsets and carries out the instruction implicit in that subset, by taking the hoop to the person or to the underwater speaker.

These different types of responses revealed a rather remarkable and intelligent analysis of the sequences. Thus, the dolphin did not terminate her response when an anomalous initial sequence such as person speaker was first detected. Instead, she continued to process the entire sequence, apparently searching backward and forward for proper grammatical structures as well as proper semantic relationships, until she found something she could act on, or not. This analytic type of sequence processing is part and parcel of sentence processing by human listeners.

Understanding of symbolic references to absent objects. Louis Herman and Paul Forestell tested the dolphin Akeakamai’s understanding of symbolic
references to objects that were not present in the dolphin’s habitat at the time the reference was made. For this purpose, they constructed a new syntactic frame consisting of an object name followed by a gestural sign glossed as “Question”. For example, the two-item gestural sequence glossed as basket question asks whether a basket is present in the dolphin’s habitat. The dolphin could respond Yes by pressing a paddle to her right or No by pressing a paddle to her left. Over a series of such questions, with the particular objects present being changed over blocks of trials, the dolphin was as accurate at reporting that a named object was absent as she was at reporting that it was present. These results gave a clear indication that the gestures assigned to objects were understood referentially by the dolphin, i.e., that the gestures acted as symbolic references to those objects.

Interpreting language instructions given through television displays. The television medium can display scenes that are representations of the real world, or sometimes of imagined worlds. As viewers, we understand this and often respond to the displayed content similarly to how we might respond to

the real world. We of course understand that it is a representation, and not the real world. It appears, however, that an appreciation of television as a representation of the real world does not come easily to animals, even to apes. Sue Savage-Rumbaugh wrote in her book, Ape Language, that chimpanzees show at most a fleeting interest in television, and that from their behavior it was not possible to infer that they were seeing anything more than changing patterns or forms. Her own language -trained chimpanzee subjects, Sherman and Austin, only learned to attend to and interpret television scenes after months of exposure in the presence of human companions who reacted to the scenes by exclaiming or vocalizing at appropriate times. Louis Herman, Palmer Morrel-Samuels and Adam Pack tested whether the dolphin Akeakamai might respond appropriately to language instructions delivered by a trainer whose image was presented on a television screen. Akeakamai had never been exposed to television of any sort previously. Then, for the first time, the researchers simply placed a television monitor behind one of the underwater windows in the dolphin’s habitat and directed Akeakamai to swim down to the window. On arriving there she saw an image of the trainer on the screen. The trainer then proceeded to give Akeakamai instructions through the familiar gestural language. The dolphin watched and then turned and carried out the first instruction correctly and also responded correctly to 11 of 13 additional gestural instructions given her at that same testing session. In further tests, Akeakamai was able to respond accurately even to degraded images of the trainer, consisting, for example, of a pair of white hands moving about in black space. The overall results suggested that Akeakamai spontaneously processed the television displays as representations of the gestural language she had been exposed to live for many years previously.


The results of the language comprehension work with the bonobo chimpanzee and the dolphin Akeakamai show many similarities, especially in the receptivity of the animals to the language formats used and in their proficiency at responding to sequences of symbols. The dolphin has been tested in more formal procedures than has the bonobo, leading to a fuller understanding of the dolphin’s grammatical competencies than has been attained for the chimp. The findings with the bottlenosed dolphin are in keeping with many other demonstrations of the cognitive abilities of this species. The advanced cognitive abilities of apes are also well documented. An early summary by Herman (1980, p. 421) still seems appropriate to accommodate the convergent cognitive and language-learning abilities of ape and dolphin: “The major link that cognitively connects the otherwise evolutionarily divergent (dolphins)... and primates may be social pressure--the requirement for integration into a social order having an extensive communication matrix for promoting the well-being and survival of individuals…. Effective functioning in such a society demands extensive socialization and learning. The extended maturational stages of the young primate or dolphin and the close attention given it by adults and peers…provide the time and tutoring necessary for meeting these demands. In general, high levels of parental care and high degrees of cortical encephalization go together…. It is not difficult to imagine that the extensive development of the brain in (dolphins)…and the resulting cognitive skills of some members of this group, have derived from the demands of social living, including both cooperation and competition among peers, expressed within the context of the protracted development of the young. These cognitive skills may in turn provide the behavioral flexibility that has allowed the diverse family of (dolphins)…to successfully invade so many different aquatic habitats and niches.”


Caldwell, M. C. and Caldwell, D. K. (1965). Individualized whistle contours in bottlenosed dolphins (Tursiops truncatus). Nature (London), 207, 434-435.

Chomsky, N. (1972). Language and mind. Harcourt Brace Jovanovich, New York.

Gardner, B. T. & Gardner, R. A. (1971). Two-way communication with an infant chimpanzee. In A. M. Schrier & F. Stollnitz (Eds.). Behavior of nonhuman primates. Vol. 4. pp. 117-184. Academic Press, New York:.

Herman, L. M. (1980). Cognitive characteristics of dolphins. In Cetacean behavior: Mechanisms and functions (L. M. Herman ed.) pp. 363-429. Wiley Interscience, New York.

Herman, L. M. (1986). Cognition and language competencies of bottlenosed dolphins. In Dolphin cognition and behavior: A comparative approach (R. J. Schusterman, J. Thomas, and F. G. Wood eds.) pp. 221-251. Lawrence Erlbaum Associates, Hillsdale, NJ.

Herman, L. M. (1989). In which Procrustean bed does the sea lion sleep tonight? The Psychological Record, 39, 19-50.

Herman, L. M. and Forestell, P. H. (1985). Reporting presence or absence of named objects by a language-trained dolphin. Neuroscience and Biobehavioral Reviews, 9, 667-691.

Herman, L. M. and Tavolga, W. N. (1980). The communication systems of cetaceans. In Cetacean behavior: Mechanisms and functions (L. M. Herman, ed.) pp. 149-209. Wiley Interscience, New York.

Herman, L. M. & Uyeyama, R. K. (1999). The dolphin’s grammatical competency: Comments on Kako (1998). Animal Learning & Behavior 27, 18-23.

Herman, L.M., Kuczaj, S. III, & Holder, M. D. (1993). Responses to anomalous gestural sequences by a language-trained dolphin: Evidence for processing of semantic relations and syntactic information. Journal of Experimental Psychology: General 122, 184-194.

Herman, L. M., Morrel-Samuels, P. and Pack, A. A. (1990). Bottlenosed dolphin and human recognition of veridical and degraded video displays of an artificial gestural language. Journal of Experimental Psychology: General, 119, 215-230.

Herman, L. M., Richards, D. G. & Wolz, J. P. (1984). Comprehension of sentences by bottlenosed dolphins. Cognition, 16, 129-219.

King, B. J. and Shanker, S. G. (1997). The expulsion of primates from the garden of language. In Evolution of Communication (S. Wilcox, B. King, and L. Steels, eds.) Vol. 1, Number 1, pp. 59-99. John Benjamins Publishing Company, Philadelphia.

Lilly, J. C. (1967). The Mind of the Dolphin: A Nonhuman Intelligence. Doubleday, New York.

Premack, D. and Premack, A. (1983). The Mind of an Ape. W.W. Norton & Company, Ltd., New York.

Richards, D. G., Wolz, J. P. & Herman, L. M. (1984). Vocal mimicry of computer generated sounds and vocal labeling of objects by a bottlenosed dolphin, Tursiops truncatus. Journal of Comparative Psychology, 98, 10-28.

Ristau, C. A. and Robbins, D. (1979). Language in the great apes: A critical review. In Advances in the Study of Behavior (J. F. Rosenblatt, R. B. Hinde, C. Beer and M-C Busnel, eds.) Vol. 12, pp. 141-255. Academic Press, New York.

Rumbaugh, D. M. (1977). Language Learning by a Chimpanzee: The Lana Project. Academic Press, New York.

Savage-Rumbaugh, E. S. (1986). Ape Language: from Conditioned Response to Symbol. Columbia University Press, New York.

Savage-Rumbaugh, E. S., Murphy, J., Sevcik, R. A., Brakke, K. E., Williams, S. L., and Rumbaugh, D. M. (1993). Language comprehension in ape and child. Monographs of the Society for Research in Child Development 58, No. 3-4.

Schusterman, R. J. & Krieger, K. (1984). California sea lions are capable of semantic comprehension. The Psychological Record, 38, 311-348.

Schusterman, R. J. & Gisiner, R. (1988). Artificial language comprehension in dolphins and sea lions: The essential cognitive skills. The Psychological Record, 34, 3-23.

Terrace, H. S., Petitto, L .A., Sanders, R. J. & Bever, T. G. (1979). Can an ape create a sentence? Science, 206, 891-902.

Tyack, P. L. (1986). Whistle repertoires of two bottlenosed dolphins, Tursiops truncatus: mimicry of signature whistles? Behavioral Ecology and Sociobiology, 18, 251-257.

Back to Top

Dolphin Programs | Whale Programs | Education Programs | Our Research | Resource Guide

Copyright © 2002, The Dolphin Institute