Dolphin Mental Abilities
II. EXPERIMENTS with DORIS and BUZZ
[Continued from Part 2]
In 1964, another interesting experiment in abstract dolphin communication began. The experiment encouraged a three year old female dolphin and a four year old male dolphin to devise their own abstract and arbitrary acoustic solution to an experimental problem. The designer of the experiment, Jarvis Bastian, thought that the dolphins might be able to indicate the use of dolphin "natural" language or at least the equivalent of a word in solving the task. This experiment was called "The Transmission of Arbitrary Environmental Information Between Bottlenosed Dolphins."
There were two phases to the experiment: Part I, 1964-1965 (Bastian, 1966); Part II, July 1966 to July 1968 (Bastian, Wall and Anderson, "Further Investigation of the Transmission of Arbitrary Environmental Information Between Bottlenosed Dolphins", 1968).
The first phase was presented at a symposium in Italy on Animal Sonar Systems and published in the symposium book. The second phase was never completed, but a report is available from the Navy, some libraries or from Dr. Jarvis Bastian at the University of California at Davis.
These experiments used a complex training schedule of bridges and strict visual and sound controls to demonstrate that a female dolphin, Doris, in one half of a bisected tank, could communicate acoustically, through a visual barrier, to a male dolphin, Buzz, that a light stimulus was either on steadily or flashing. It was then up to Buzz to push the correct paddle associated with the steady or flashing light, in order for both dolphins to receive a food reward.
The first phase ended in May 1965 after thousands of trials with a 90% success rate. One point made in the second report shows a significantly questionable assumption on the part of the experimenters. The assumption was that two dolphins, caught when they were just past infancy, could already possess a suspected natural abstract communication system. These dolphin's competency in this communication system would have to be sufficient for testing the hypothesis of the experiment. On page 805 of the symposium book (page 3 of the article), under preliminary training, Bastian says;
"The two animals participating in this test were a male and female who were estimated to be 4 and 3 years old respectively at the start of this training. They had been captured about a year before, during which time they had been independently trained to accept dead fish by hand."
This makes the dolphins between 2 and 3 at the time of capture. That age is much too young to expect dolphins, whose development may closely approximate human development, to have learned a "natural" dolphin communication code, assuming one exists. A better choice of subjects would have been adult dolphins (see Zanin et al., 1990; and Markov and Ostrovskaya, 1990), not near infants. Considering the extents to which efforts have been made by Bastian, Evans, Wood, Herman, and others to show that this experiment was not a demonstration of language, it is interesting that none of them pointed out this glaring flaw in the choice of subjects.
Bastian says in the original report's conclusion that it had not been possible to establish the exact nature of Doris' emissions on which the successful performances were based. Bastian did report that analysis of the recordings appeared to isolate pulsed signals, their onset and repetition rate/duration, as probable indicators. These indicators were confirmed in the second report. It was also noted in the second report that during the stimulus reversal experiment, the female gradually stopped making pulsed sounds during the flashing light stimulus. This became another indicator, in a process started, but not completed, in the experiment.
One reason the pulsed train was able to be isolated as the female dolphin's solution to the problem is because the dolphins were sometimes trained and/or retrained separately at various stages of the experiments, and during this process the dolphins did not whistle (Bastian et al., 1968, table 1, page 7). Whistling has been identified as exclusively a socially contextual behavior. Under these conditions there was no other dolphin with whom it could communicate. Therefore, the dolphins would not be expected to whistle. Part of the confusion regarding the experiment, and the reason for the great extents to which the experiments were designed and adapted, is due to the faulty assumption mentioned earlier. The assumption that successful communication would contain the equivalent of a "word" in subjects too young to have learned such signals. But the experimental requirements were apparently simple enough that success was assured with no more than a pulsed train (an arbitrary signal). On this subject Bastian mentions that Lilly and Miller suggested in 1961 that low repetition rate pulses are used in different contexts than high repetition rate pulses. Higher rates are usually seen in echolocation and low rates, like the ones used by Buzz in Bastian's study, are usually seen in social contexts.
The closest examples of such problem solving in humans are the development of arbitrary verbal or hand-sign communications in young children left to spend a lot of time with each other. These include verbal communication in children close in age or twins (usually female), and deaf children of non-deaf parents (sign language). Such communication would greatly facilitate cooperation in play behavior. These arbitrary "languages" are usually based on the parental language. Interestingly, this untrained verbal and hand-sign approach to communication problem solving had previously not been see in non-human animals. Unfortunately, this unique aspect of the experiment is not mentioned in either report. In fact, the experimenters attempted to refute their own data. Further, there are examples of inherently faulty interpretations in the literature about these experiments.
Jarvis Bastian and Bill Evans in a later paper on "Marine Mammal Communications" (in The Biology of Marine Mammals, H. T. Anderson, ed., 1969), imply that the second phase of the experiment was completed. Evans worked for the Navy at the time of the experiments and this paper. They say;
". . . Although at the time of the original report it had not been possible to establish the exact nature of the female's acoustical emissions on which this performance success was based, subsequent analysis and further changes in the nature of her signals (as yet not published) now permit precise characterization of the critical signaling events."
The second report was not published because the research was not completed. The two playback and the role reversal experiments were not finished due to technical difficulties. Only the retraining effort and the stimulus reversal experiment were completed. The ambitious second phase to the experiment actually became only a stimulus reversal experiment. The "precise characterization of the critical signaling events" was further analysis of information from the stimulus reversal experiment that provided similar data to that of the original experiment and only confirmed the speculations listed in the conclusion of the original paper. The "precise characterization" was supposed to come from the two cancelled playback experiments.
After a "tortuous detailed account" (their words) justifying a narrow interpretation of intention in human communication (involving a mother warning her child about the approach of a truck), Evans and Bastian explain Bastian's experiment. They then conclude that the responses of the dolphins were the same as if they had been trained separately to respond to the experimental cues and therefore the dolphins did not intend to communicate, although it seems on the surface to be similar. Herein lies the dispute. Evans and Bastian say,
". . . Therefore, we may proceed in this case by asking whether we have grounds for thinking that the female in some sense knew what the consequences of her signals were for the male, apart from her consistent emissions of the necessary signals. In fact, there is no evidence at all to indicate her understanding of the relation of her emissions to the male's actions and quite a bit indicating her complete innocence. For example, when the visual barrier between the animals was withdrawn, and even when the male was removed from his enclosure, she continued to emit the signals as long as usual events ensued. The unpublished recent experimental results have now quite firmly established that the basis of this rather complex performance is very much what might have been expected had the female been trained by herself to emit signals differentially in the presence of the different visual signals and if the male had been separately trained to press his paddles differentially with respect to these acoustical signals. This is actually what seems to have happened, except that, in effect, they trained themselves."
A careful reading of the two lengthy reports shows that this statement is simply not true!
First, there is evidence that Doris intended to communicate with Buzz. During part of the retraining Doris and Buzz were retrained separately (Buzz was not reported in either paper to have been removed from the tank during any of the experiments). During this separated phase, Doris responded correctly, but the data noted that she made no whistle sounds (session A, table 1,, Bastian et al., 1968). When Buzz was reintroduced later, Doris resumed her whistling. Since many researchers, including Evans, have expressed the social communications value of whistling among dolphins (Dreher, 1961; Dreher and Evans, 1962; Evans and Prescott, 1962; Caldwell and Caldwell, 1965, 1968), it can be assumed that her resumption of whistling was an "indication" demonstrating her "intent" to communicate, yet Evans and Bastian do not mention this information.
Second, when the visual barrier was removed, it was never removed completely as their statement implies, and the dolphin may not have been able to see each other during this stage. The barrier was only partially drawn back and just enough, by the judgment of the experimenters, to maintain a 90% level of success during the gradual closing process. To do otherwise would have been a waste of time and the second report explains that when they stopped the research, they were behind schedule.
Third, Bastian and Evans neglect to say that for the "usual events to ensue" when the dolphins were separated, the experimenters must take the place of Buzz and/or Doris as mentioned on pages 7, 17 and 38 of the second report. This had to be done so that the dolphin's correct sounds or actions could be reinforced with fish. By design, working for food is the reason why they had to cooperate with each other, or the experimenters, in the first place. Without experimenter cooperation (playing the part of Buzz), Doris would have stopped making pulsed trains. Psychologists call this "extinguishing a response. If Doris received no fish for her sounds, she would have stopped making them, thus Evans and Bastian would have had no "indications" and therefore evidence of lack of "intent". It would have been better for Evans and Bastian to have claimed that the dolphins intent was to acquire food rather than communicate, but that would still not eliminate the possibility that they intended to acquire food by using abstract communication.
Finally, the last part of the statement shows how Evans and Bastian chose to take their interpretation of the experiment out of context in order to make their point. The fact that the dolphins were indeed exposed to each others responsibilities in order to get fed; the fact that they did solve the problem by training themselves, (most likely without normal dolphin communication code, as evidenced by their young age and shown by the aborted role reversal experiment); and the fact that the data of both experiments back up this interpretation; are strong indications that the dolphins intended to cooperate and communicate in order to get fed. How would Evans and Bastian interpret the performance of pre-linguistic 2 or 3 year old human children in the same experiment? They may solve the task as the dolphins did, but would they have intended to solve the task.
It has been shown that Evans and Bastian fail the test of showing lack of "intent" on the part of Doris and Buzz when the original data is reviewed. Their attempts to draw conclusions by looking at selective parts of the data, ignoring behavioral sequences shown by the data, and finally, ignoring results registered in the data, but required by the interpretation they have chosen, is highly significant and problematic. These points remain unchallenged to date.
Their justification for making this out-of-context interpretation is stated at the end of this section in the article. Evans and Bastian say,
". . . We hope that a full appreciation of (this interpretation) will free inquiry into, and interpretation of, the social life of these animals from those aspects of their social behavior that parallel some of the attributes of man's linguistic interactions. One of our foremost goals in what follows is to sensitize this are of research to the fact that much more than vocal-auditory pathways are involved in the investigation of the social lives of these animals." (parenthesis information added)
It seems that Evans and Bastian wish to encourage research into dolphin social life, but think that researchers should not call anything marine mammals (especially dolphins) do vocally, linguistic. In this respect they would like to re-direct dolphin research away from concepts that may very well be at the root of why the Navy classified its dolphin program. What happens, though, when they are wrong and the vocal behavior of dolphins is shown to be as complex as human language? After all there is no scientific information available that shows that dolphins do not communicate abstractly using whistles. Is it language? Only research will tell and Evans and Bastian weigh in against further research which is very unscientific. Please remember that this research was done under Navy contract. The first part was conducted prior to classification of the Navy's dolphin program. The second part, where scientific protocol and factual reporting were disregarded was conducted after the program was classified.
Glasnost and the fall of the Soviet Union has allowed out of Russia two articles that shed light on these developments. They are a partial replication of Bastian's experiment by Zanin, Markov and Sidorova called "The Ability of Bottlenose Dolphins, Tursiops truncatus, to Report Arbitrary Information" (1990), and Markov and Ostrovskaya's "Organization of Communication System in Tursiops truncatus Montague" (1990). The articles read in English like someone with a heavy Russian accent. They are quite interesting as they indirectly reveal the mistakes of Navy researchers, such as using dolphins caught too young to have learned dolphin communication code in Bastian's "language" experiment.
The first paper uses adult female dolphins in a experimental paradigm modeled closely but not exactly after Bastian's. Aside from the use of adults, their major modification replaced light signals with balls of different size and incorporated role reversals successfully into the same experiment. In Bastian's experiment, the role reversal segment was not completed, but enough data was accumulated, before they shut down the experiment, to show that there was not an easy exchange of roles and thus probably no "words" that were easily used by both dolphins. The Russian experiment was not as ambitious as Bastian's nor was the data analyzed as extensively (because the Russians were not trying to prove a lack of language). But the dolphins did cooperate in "joint work" with a high percentage (83%) and in 90% of all the joint work pressed the same "manipulandum". After the high percentage of Session A, though, the response percentages began to decline until the dolphins refused to "talk" to each other in sessions C, D and E. This breakdown in communication was due to one of the dolphin's (Jenny) dominance over the other (Kora) and her actions to stop the role reversals (where Kora would tell Jenny which ball to press). Zanin et al. finish the article with;
"Comparing the experimental data which were obtained in the sessions, with the results of the theoretical calculations, we conclude that the interchange by (sic - of) the information between the dolphins was observed in the session 'A', and to a lesser extent in session 'B'; the interchange was absent in the sessions 'C', 'D' and 'E'. The results of the observation on the acoustic activity of the dolphins and the data about their behavior in the sessions coincide: the acoustic activity, which was maximum in the session 'A', became less in the session 'B' and
disappeared completely in the next session.
"Thus, our results show that bottlenose dolphins can report arbitrary information and can co-ordinate the behavior of each other. . . ."
The relative ease with which the dolphins reversed roles would mean, if we are to use Bastian et al.'s reasoning, that the dolphins were using an existing equivalent of a "word" from their own adult dolphin communication code.
The second article shows that dolphin communication research analyzed using linguistic tools reveals a communication system among dolphins as complex as human language.
In this analysis Markov and Ostrovskaya say,
"Since signal structure is generated in time, a signal is a chain of differently sounding blocks (irrespective of their complexity). The number of such blocks in a signal varies from 1 to 24, averaging 5-7. The number of structural types of blocks has not been established definitely, but it is well over one hundred. Using this data and standard formulas from Games Theory, one can calculate easily that 10 to the 12th power signals could be produced by means of free combining. True, it is a potential estimate, but even if we assume that, because of code-associated or physical block combination bans, only one ten-millionth of them actually is used, still there remain 10 to the 5th power signals available for communication, which certainly is more than required for actual communication. All this makes it possible to think that the communicative system of bottlenose dolphin is 'open' in terms of vocabulary formation. This conclusion is indirectly supported by the fact that dolphins use hundreds of structural types of signals for communication (see, for instance, Table 3)."
This article outlines four parameters influencing the complexity of dolphin communication code.
- 1. Juvenile dolphins use stereotypic signals;
- 2. stressed dolphins make stereotype signals;
- 3. dolphins in separate tanks connected by "telephones", with bandwidths that restrict the frequencies normally used by the dolphins, make stereotypic signals;
- 4. adult dolphins that are calm and adapted to their surroundings communicate with a code as complex as human language when this code is analyzed using linguistic tools.
The first three conditions represent the failings of three types of American experiment used to show that dolphin communication is stereotypic and not complex or open.
After the linguistic analysis, Markov and Ostrovskaya conclude,
"Summing-up data provided in the paper,, one concludes that bottlenose dolphins have an open-type communicative system. Using a very flexible sound generation system, they can use multi-level combining for constructing a virtually unlimited vocabulary - a set of acoustic signals - for forming a multitude of organized text messages out of its units. As to degree of complexity, the communicative system of bottlenose dolphins is unique, and nothing of the kind has so far been discovered in the other animal species."
Two of the three types of failings mentioned in this article have been indicative of oceanarium sponsored communication research, often with Navy cooperation or funding. The two failings are:
- 1. expecting babies and juveniles to have adult dolphin communication abilities even after they grow up without ever being exposed, for any reasonable length of time, to a dolphin caught as an adult; and
- 2. recording stressed dolphins to see if dolphins have "language".
- 3. using "telephones" with too small a bandwidth, was a single Navy experiment which happened to use Buzz and another dolphin as subjects. This is the same Buzz from Bastian's experiment.
Presented by "Whales in Danger" - Whales on the Net - http://www.whales.org.au
[Continue to III. - EXPERIMENTS with MALIA and HOU]
Whales in Danger Information Service