Dolphin Mental Abilities Paper

by Kenneth W. LeVasseur <>


An experiment which appears to have been extremely frustrating for the subject (Hou in this instance) was Karen Pryor's "The Creative Porpoise: Training for Novel Behavior". This experiment was done in the mid-sixties at Sea Life Park in Hawaii, and was published in 1969. It concerns the creative abilities of two rough-toothed dolphins (Steno bredanensis), one named Malia (a regular show performer at Sea Life Park), and one named Hou (a research subject at adjacent Oceanic Institute). This creative ability entails the abstract concept of consistent inconsistency, or the ability to generalize, without shaping (the means by which trainers form a new behavior by reinforcing close approximations to the desired behavior) as a training tool, that the trainer wants demonstrations of new (unnatural, novel, untrained) behavior. It took only a few days of continuously repeating the first step in conditioning (reinforcing new or more specific behaviors) for Malia to emit an unprecedented range of behaviors. By the fifteenth session, it had become apparent that Malia not only knew that only new behavior would be reinforced, but also that only one behavior per session would be reinforced. Malia went on show after show producing new and astonishing responses. "Furthermore, the trainers could not imagine shaped behaviors as unusual as some emitted spontaneously by the porpoise." (Pryor et al., 1969).

This new application of the first step in conditioning provided inspiration for an experiment that would pinpoint the moment of originality, what some would call "point of insight" in the dolphin's performance. Unfortunately, the experimental method was compromised, as acknowledged by Pryor in the article on the experiment. But, when the experiment is placed in context, some information is retrievable. The technological applications, though, were not lost, as Karen Pryor says in her book, Lads Before the Wind (1975):

". . . I was not surprised to learn that Navy trainers developed a technique, based on this experiment, of both refreshing an animal and extending its awareness and sophistication by having 'playtime' between or after more stringent training sessions, in which all kinds of things would be reinforced, and from which, sometimes, came new and useful responses."

The major scientific criticism of the experiment with Hou stems from the methodological error of introducing the second step of conditioning, shaping, into an experiment designed to test the production of behavior without shaping. Pryor says:

"The experimental plan of reinforcing a new type of response in each session was not fully met. Sometimes a previously reinforced response was again chosen for reinforcement. . . . Whether the 'reviewing' of responses was helpful (emphasis added) or detrimental to the animal's progress is open to speculation."

The reasons for interrupting the planned procedure were varied. The initial one came in the first few sessions in which there was a photographic review planned into the schedule for a Navy documentary on the experiment. Since the Navy was paying for the experiment they had the "right" to film its procedures and results. But, in order to review previous behaviors and patterns, these had to be shaped; and the introduction of shaping into the critical early sessions of the experiment would have a high probability of confusing the subject. This confusion can be expected to interfere with Hou's understanding of the nature of the experimental task. Furthermore, this critical interruption is not completely reported in the journal article on the experiment (Pryor et al., 1969), where only brief mention is given to the "filming". There was also no need to interrupt the experimental procedure, because the cameraman could have just filmed the behaviors that were being produced as part of the experiment. By distorting the experimental procedure, through shaping for filming purposes, the question of authenticity in the film is raised. This excuse for the initial interruption does not appear to be a valid form of reasoning. Other reasons in the report for introducing shaping were to strengthen the response; to increase the general level of responding; to interrupt unvarying repetition of a limited repertoire; and to prevent a low level of reinforcement from leading to extinction of all responses. Pryor says:

"While this represented a departure from the primary goal of conditioning novel behavior, the experimenters realized that Malia, the show animal, had experienced some training sessions in which, no new spontaneous action being emitted, some specific response was shaped. It was not known whether or not the shaping sessions had contributed to Malia's ability to emit novel responses. Therefore, the inclusion of shaping in Hou's training seemed permissible."

The possibly beneficial interruptions of Malia's experiment have an implied justification, albeit unverified. The extensive, and obvious unjustified, interruptions of Hou's experiment can easily be seen as feeding Hou's confusion, especially in light of the fact that the original introduction of shaping for filming purposes was introduced before Hou had any problems. In fact, the nature of the reasoning for the continued interruptions appear to be compensation for the original interruption for filming purposes. Comparing Hou's interruptions with Malia's reviewing is also misleading. It is stated that Malia was a show dolphin. The experiment in training that Pryor turned into a show was, after all, still a show. In a show there is pressure to produce a performance. People will become bored if they have to sit around while a dolphin has to think about the solution to a problem, especially if there is no guarantee that the dolphin will solve the problem and produce a performance. There was performance pressure to get Malia to solve the problem. There was no such pressure in the Hou experiment because there was no show. Pryor had the chance to show scientifically, by following the proper procedure, whether the dolphin could solve the problem without reviewing and without the pressure of performance. The primary scientific goal of the experiment was compromised for a filming schedule as defined by the funding agency the U. S. Navy.

Many examples are given in the film, the scientific paper and Pryor's book, which show the confusing results of these extensive and scientifically unjustifiable interruptions. Considering the interruptions of the experiment, it is a wonder that Hou did not take longer to solve the task or that Hou succeeded at all. A simple comparison of the performance of the two dolphins suggests the severity of the interruptions. The obviously contraindicated but hopeful nature of Pryor's suggestion that the interruptions could have had a "helpful" effect on Hou's experiment is misleading. The effect was obviously detrimental, Malia solved the task in 15 sessions while Hou took more than twice as long solving a similar task in 33 sessions. Pryor should have stated this obvious point, if only to show how the two procedures were different.

An example of how these comments can affect interpretations of dolphin behavior is seen in Herman's 1980 book Cetacean Behavior. On pages 400-401, while describing conceptual processes (specifically "Forming a Concept About Motor Responses") in the chapter Cognitive Characteristics of Dolphins, Herman describes this experiment saying

". . . Approximately 16 sessions were required for new behaviors to emerge with some regularity, although older previously-reinforced behaviors were more common. After the 16th session, the number and percentage of new behaviors offered began to increase substantially, and at the end of an additional 17 sessions the repertoire of behaviors had become so enlarged that the experimenters had difficulty in reliably classifying an ongoing behavior as old or new. The experiment was ended at that point."

In his comments Herman does not mention that the performance of Hou was interrupted numerous times, probably leading to confusion and the unusually high number of sessions, nor doe he point out the broad discrepancy on this point between Hou's and Malia's performances. Considering the thoroughness of Herman's book and its popularity among students of dolphin and other animal behavior, the perpetuation of inaccuracies concerning Hou's performance is regrettable.

Oddly enough, when Pryor's students repeated the planned procedure with a pigeon, they did not similarly interrupt the procedure. But, Pryor was willing to suggest that the "same" procedure was used and that the pigeon results were a valid comparison with Hou's experiment - an obvious misuse of the data.

The only scientific comparison available between the pigeon experiments and the dolphin demonstrations (which Pryor does not attempt to use) is that novel behavior was produced by different but vaguely related procedures that depend on the emotional and mental abilities of the subjects to cope with interruptions of those procedures. It is currently well known that novel behavior (originality) occurs in a wide range of species and that such novel behavior can be encouraged with the classical conditioning of current training techniques. This is achieved by using the first step in conditioning (bringing behavior under signal 'stimulus' control) and then only reinforcing new responses. The major differences which may exist between species in this type of experiment are that the success of the procedure could depend on the other mental abilities of the different species. These mental abilities include; how much novel behavior can be produced before the subject shows frustration (Maltzman, 1960) or at what optimal rates novel behavior can be solicited from each species. For example, in Malia's and Hou's demonstrations, there were five sessions per day (five novel behavior patterns), yet in the pigeon demonstration mentioned by Pryor, only one behavior pattern was attempted and reinforced per day, and the pigeon was only able to provide a limited set of novel responses although they were "quite unusual".

Related Research

Similar differences can be seen in the varying rates and/or number of trials needed before selected species can be brough under signal (stimulus) control during training (learning sets). In a paper on learning sets in dolphins, Herman and Arbeit (1973) point out that:

"For most mammalian species tested, the probability of a correct response rises over successive blocks of problems, but the rate of increase and the final probability level reached vary widely among species... Performance levels on DLS (Discriminative Learning Set) tasks across species are generally in keeping with an ordering base on cortical complexity (Harlow, 1959). The highly convoluted surface and large size of the brain of the Atlantic bottlenosed dolphin (Flanigan, 1972; Jansen and Jansen 1969) suggests the capacity for a high degree of learning efficiency . . ."

"The trial 2 performance of the dolphin, 86 to 100% correct responses . . . equaled or exceeded average Trial 2 levels observed for higher primate species in visual DLS tasks (Devine, 1970; Harlow, 1949; Miles, 1965; Riopelle and Chinn, 1961; Rumbaugh and McCormack, 1967; Warren, 1965). Runs on errorless problems in some cases rivaled those reported for individual rhesus monkey subjects (Harlow, 1949). This suggests a revision to Warren's (1965) statement that 'no nonprimate mammal yet tested has approached the level of one trial learning (in DLS tasks) observed in primate species (p. 266)'."

Thompson and Herman (1977) tested the memory of the bottlenosed dolphin for lists of sounds in which the memory data bore many similarities to data obtained from humans tested on probe recognition tasks. Specifically they say,

"Overall, our results for the dolphin were similar to results obtained from human subjects in probe recognition tasks and revealed many of the same capabilities and constraints observed in human prformance."

The paper ends with:

"The human data are extensive and complex and we have here described only some broad functional similarities to it. Nevertheless, obtaining these similarities within a paradigm closely following that used with humans encourages the view that we may be dealing with much the same underlying conceptual phenomena discussed in human memory work, implying a convergence of some memory control process in dolphin and human, both highly sophisticated species acoustically."

Further work on the same experiment in which the lists of sounds were extended from 6 to 8 sounds indicate that the Atlantic bottlenosed dolphin's strategies for remembering lists of sounds is consistent with decision strategies adopted by human subjects tested on serial probe recognition tasks (Herman, 1980 - referencing Thompson and Herman's further unpublished research).

This memory for serial lists of sounds can be conceptually expanded to cover memory for serial lists. Because each elicited behavior by Malia and Hou becomes a remembered response which must be avoided for Malia or Hou to provide "novel" behavior, this memory for serials could be critical to determining the limit of novel behavior.

Pryor et al., 1969, quote Maltzman (1960) who did a related experiment with humans by saying in their paper;

"It is interesting to note that behavior considered by the authors to indicate anger in the porpoise was observed under similar circumstances in human subjects by Maltzman:
'An impression gained from observing in the experimental situation is that repeated evocation of different responses to the same stimuli becomes quite frustrating. Subjects are disturbed by what quickly becomes a surprisingly difficult task. This disturbed behavior indicates that the procedure may not be trivial and does approximate a non-laboratory situation involving originality or inventiveness, with its frequent concomitant frustration.'

Maltzman's procedure involved the repeated presentation of a list of stimulus words in a modified free-association situation accompanied by instructions to give a different response to each stimulus.

"Under these conditions the responses became more uncommon"

Further review of Maltzman's paper reveals that misleading interruptions were not allowed in the experimental procedure.

Another insight into the experiment comes from Pryor's paper stating the reasons for terminating the testing:

"After 32 training sessions, the topography of Hou's aerial behaviors became so complex that, while undoubtedly novel, the behaviors exceeded the powers of the observers to discriminate and describe them. This breakdown in observer reliability was one factor in the termination of the experiment."

Yet Hou was limited to providing a reinforceable act which consisted of any movement that according to Pryor et al.:

". . . was not part of normal swimming action and which was sufficiently extended in space and time to be reported by two or more observers. Such behavior as eyerolling, inaudible whistling and gradual changes in direction may have occurred, but they could not be distinguished by the trainers and therefore could not be reinforced except coincidentally. This unavoidable contingency probably had the effect of increasing the incidence of gross motor responses."

Even with all the human limitations placed upon them, Malia and Hou were able to demonstrate a phenomenal creative ability. The experiment, as reported in the different mediums, does not present an accurate picture of the dolphin's creative ability or a true measure of the difficulty inherent in their performance.

Recent Language Research with Dolphins

Is there any more light that Louis Herman's dolphin language research can shed on the question of the dolphin's ability to communicate abstractly? There is! Unfortunately, Herman does not allow the dolphins in his experiments to communicate back except by their gross motor actions and pressing yes/no paddles. In this respect his methodology could easily be adapted to primate communication research where chimps and gorillas are unable to vocalize with anywhere near the flexibility show by dolphins. The ape researchers, though, prefer American Sign Language and computers for their research, possibly because they allow faster and more precise responses.

Since 1979, Herman voiced the desire (mostly in the media) to allow the dolphins in his experiments to participate in two-way communication, but has never done so. Although the dolphins in his experiments demonstrate linguistic comprehension, they are not allowed to produce language. Herman states in a 1986 article (referencing comprehension problems in apes):

". . . production does not necessarily, perhaps usually in the case of the apes, imply comprehension. For the dolphins, one might also caution that an ability to comprehend does not necessarily imply that the same animal could produce language. It may yet be shown that receptive and productive processes are sufficiently distinct for animals that transfer from one mode to the other is difficult at best . . ."

In the next paragraph, concerning object labeling and reporting, Herman once again mentions the work of Douglas Richards, done at Herman's lab (Richards et. al. 1984, Richards, 1986) where the dolphins demonstrated that they could mimic whistles and label objects, but calls it "modest first steps into the exploration of production." The sounds used in his experiments do include some FM signals (whistles), but the nature of many of the sounds Herman uses are difficult for dolphin to reproduce. He then concludes the paragraph with:

" . . . In the future we hope to use the reporting procedure to pose binary questions to the dolphin about more abstract topics than the objects to which the signs refer . . ."

Still no concrete plans for language production, even though he conceded earlier that it is an unresolved question. It certainly is an important question. At this point in the funding of Herman's laboratory (1985, when this paper was written for publication in 1986) there is a change from National Science Foundation funding to the office of Naval Research. In 1987, with funding from the Navy, Herman turns further away from eventually studying productive competencies in his dolphin subjects.

In Herman's 1987 paper "Receptive Competencies of Language Trained Dolphins", Herman quotes criticism from Premack about Herman's lack of an effort for language production in his experiments,

". .. According to Premack (1986),'Herman et al. have not yet arranged for the dolphin to produce language, and they would make a virtue of this weakness, claiming unusual advantages for comprehension. One might grant part of the claim if they would acknowledge that, however great the virtues of comprehension, adding the study of production does not detract from them' (P. 17)."

Herman then acknowledges, "Granted, of course." and quotes his earlier work that concedes the need for "expanded research" into productive competencies "relative to receptive language". Yet, Herman (with his Navy funding) still has not embarked on language production experiments. As recently as 1990, Herman is still emphasizing the superiority of his receptive competencies research over productive competencies in demonstrating the dolphins language abilities, even though it is only one part of a two part question which he had earlier acknowledged.

In a paper written with Palmer Morrel-Samuels (1990) titled "Knowledge Acquisition and Asymmetry Between Language Comprehension and Production: Dolphins and Apes as General Models for Animals" they finish up the main text by saying in the last paragraph,

"These results (with dolphins), together with those presented earlier on responses to absent objects and to lexical and semantic anomalies (with dolphins and apes), provide converging lines of evidence for the referential function of the signs used in the language. More generally, the examination of the receptive skills of dolphins has revealed capabilities for semantic and syntactic processing, and suggests that the dolphins utilize a rich network of mental representations when responding to language-mediated tasks. That such demonstrations have proved rather elusive in studies emphasizing productive skills (with apes) reflects the difficulty of obtaining objective, replicable data in such studies and, more importantly, supports the hypothesis of an attenuation of productive skills relative to receptive skills." (Parenthesis information added)

Here Herman and Morrel-Samuels are obviously discussing the advantages of receptive competencies, but the last part of the last sentence is not accurate beause the "attenuation of productive skills relative to receptive skills" has not been shown in dolphins, yet the implication (by not mentioning apes or dolphins, specifically, throughout the paragraph) is that he is discussing dolphins along with the ape research. This statement has only been demonstrated in primate language research including human language research.

In their following conclusion they state;

"The language comprehension work with dolphins, the recent work by Savage-Rumbaugh with the pygmy chimpanze Kanzi, and some of the earlier work of Premack, have revealed competencies for language not realized in other work emphasizing language production. The message from these findings is twofold:

Herman and Morrel-Samuels are here making an indirect claim about dolphin language production by associating Herman's research with the primate research which has included language production. This claim has no basis in fact. They are assuming that because all animals tested, including humans, have show better performance in receptive competencies than in productive competencies, then dolphins must also fit that mold. They are ignoring the fact that dolphins have been evolving with human size brains for many millions of years longer than any of the research subjects to which he refers, including the human language research.

Markov and Ostrovskaya in 1990 observed that dolphin's have a communication code that may be as complex as human communication or possibly even more complex. Herman's own research has yet to find linguistic limits for the dolphin subjects in his program. It could well be that, as recent users of abstract communication, primates have limited abilities for productive competencies. On a purely speculative (but enlightening) level and assuming that hominids have been using abstract forms of communication for 4 million years (an acceptable estimate considering primate communication research vs. the estimated time since the divergence of our respective species); it may be that evolution and abstract communication do not favor the highest of primates. It may be that once an order evolves using abstract forms of communication over time periods three times longer than primates (i.e. only 12 million years), productive competencies balance out receptive competencies. This scientific question has not been answered and leaves more than ten million additional years for cetaceans to evolve better forms of communication and communication abilities. This added period is still more than two times longer than primates have been assumed, by this thought experiment, to be using that form of communication. In a context such as this, Herman's and Morrel-Samuels' comments show their speculative bias and the inaccuracy of their conclusion. Their comments must be backed up with actual - not implied - productive competencies research.

On another note, a careful reading of Herman and his associates reported methods and comparison of their strategies with the primate research show that the length of time it takes for the Hawaii lab's experimental results are due to their program designs, not the mental abilities of the dolphins. The length of time it takes for dolphins to respond with gross body movements to touch the proper paddles or retrieve the proper object in an experiment are exponential compared to the vocal or gestural responses of human and ape research. Considering that Herman still has not found upper limits to dolphin mental abilities (in fact his work has been instrumental in removing the limits imposed by earlier Navy dolphin researchers), the only limits that could be misinterpreted as limits of dolphin mental abilities appear to be the products of his approach. These include the use of the dolphins gross body movements for responses, paddle presses for test responses and Herman's reluctance to allow dolphins to produce language.

With this time consuming methodology Herman has developed a reputation for precise research and it usually shows in his writing. But the concluding statements of his and Morrel-Sammuels' chapter are very unlike Herman's style, where he usually says apes when he is talking about apes and usually says dolphins when he is talking about dolphins. Using such unusual behavior to support his position of not letting dolphins "talk" back shows that he is stretching his position precariously.

The literature includes criticism of the approach limitations involved in using only receptive competencies such as that in Herman's classical conditioning form of training. Sue Savage-Rumbaugh and Brakke (1990), discuss principles that cut through Herman's defenses.

In this article they comment about Herman and others:

". . . unlike the child, all of the actions carried out by Ake, Rocky, and the others in the test paradigm achieve a single goal--the receipt of a fish to eat (Herman et al. 1984; Herman 1987,, 1988; Schusterman & Krieger 1984, 1988). In the routines of daily life, a child learns to carry out different actions for different ends and becomes a functional partner in the interactions. The correspondence between the communication and its result, that is natural for the child, is not present in the dolphin and sea lion studies. . . .

" . . . The dolphin, however, has little reason to learn in this way, since it

It is more interested in reinstating the 'effect' of receiving a fish than in the actions that were the result of the symbols 'take the frisbee to the hoop.'

". . . This is not to say that these animals lack intentionality and do not communicate to each other or to humans. Within the test situation, however. these capacities have no opportunity for expression. . . .

Children learn many words that permit them to affect their world in many different ways. Dolphins learn many words that affect the world all in the same way (they produce a fish). Until training paradigms are utilized that permit the animal to do more than obtain one type of reward, the subjects cannot be expected to realize their potential for symbolic communication."

Herman's resistance to language production in his dolphin research is being shown to be against his own professed philosophy and that of other cutting edge researchers. What are Herman's thoughts on Batteau's work? In 1980, Herman acknowledges about Batteau's research that,

". . . As it stands, the experiment demonstrated the feasibiiity of using artificial whistle sounds as signals for controlling complex responses of the dolphin, but, because of the design limitation noted, failed as an evaluation of lanquaqe-learning capability."

Herman has since shown there to be considerable linguistic ability on the part of the dolphin, and like Batteau, the ability of dolphins to duplicate complex whistles. The design limitation in Batteau's work that Herman writes about was,

". . . commands such as BIP or BAIEP denoted a complex of action, object, and agent. There was no attempt to represent these unique language elements with unique sounds, nor was there provision for later dissociating these command words into their component semantic elements, . . ."

This statement reflects a narrow interpretation of Batteau's aborted work and ignores Batteau's testing with abbreviated phrases. Even Herman's work is not phonetically based, nor is the dolphin and ape gestural research. Batteau's approach, on the other hand, allows Busnel's work on the phonetics of human whistled language to be applied to dolphins. This technique would allow a phonetics based form of communication research with an animal whose brain is as large or larger than the human brain. This is something that has not yet been tried publically with dolphins or any other animal.

There is still no scientific measure of upper limits to dolphin mental abilities. As of March 1992, Louis Herman has virtually eliminated acoustic communication from the language work at his Hawaii lab in favor of the gestural approach. It is hoped that the debate about dolphin mental abilities will rage on until someone demonstrates an upper limit to dolphin mental abilities. Unfortunately, there has only been significant effort to belittle dolphin mental abilities. There is no scientific interest or effort to develop a whistled language experiment with dolphins. Is it because scientists are still not prepared to find that there are no upper limits and that dolphins may have mental abilities comparable to humans.


Back to MENUBack to CONTENTS