4. Animal & Human Language
Elvar: the whistling dolphin
There is much anecdotal lore about the intelligence of dolphins and whales. Until the 1960s, though, no scientific attempt had been made to determine their communicative abilities. In one of the first studies, Lilly (1962, 1965) tried to teach a dolphin to force air through its blow-hole in such a way that it would allow the dolphin to imitate human speech sounds. A young male dolphin named Elvar produced approximations of the word ‘squirt’, which Lilly had been trying to teach him to pronounce. Lilly also claimed that
Elvar interchanged human sounds with dolphin sounds as if he were attempting to translate, but Lilly provided no scientific substantiation in this regard. In fact, pronunciation difficulties were so great that Lilly was obliged to discontinue the study. He then moved on to investigate the means by which dolphins communicate with one another.
Notwithstanding Lilly’s extravagant claims to the contrary (including claims that dolphins have an intelligence and a religion(!) that is superior to those of humans), research has yet to show that these animals use anything as complex as what we could call language.
Akeakamai and Phoenix: learning artificial languages through sight and sound
A radically different and more scientific approach to the teaching of lan- guage to dolphins was later initiated by Louis Herman at the Dolphin Insti- tute, University of Hawaii. In an early study, rather than have dolphins mimic human sounds, Herman and his associates (Herman and Wolz, 1984) trained a bottlenosed dolphin to mimic computer-generated sounds.
The dolphin not only demonstrated that it could learn to make new whis- tles but also that it could apply these whistles to the naming of objects such as ‘ball’, ‘hoop’, and ‘frisbee’. This is similar to the apes’ abilities to name objects through sign language or computer symbols. From this production- oriented research, Herman then turned his attention to the primary process of language comprehension. In his investigation of the dolphin’s language comprehension, Herman and his associates (Herman et al., 1984) conducted experiments using two different types of artificial languages, one involving sounds, the other involving visual gestures, i.e. signs. He wanted to see if, or how well, dolphins could learn to comprehend language. This was done not only to see if the animal could learn a human-created language system, but also to discover more about the cognitive abilities of dolphins.
In 1979, a teaching programme was begun with two dolphins, Phoenix and Akeakamai (the latter’s name meaning ‘lover of wisdom’ in the Hawaiian language). Each dolphin learned one of the two artificial languages. Akea was taught the gesture-based language, while Phoenix was taught the sound- based language. Each was taught a vocabulary of about 30 words, mainly names of objects, agents, actions, and modifiers. The sound-based language had its sounds projected underwater into the dolphin tank. These sounds were controlled by Herman and his assistants from their underwater labor- atory, which had a window view into the tank.
The visual language of gestures, invented by Herman and his colleagues, involved the use of the trainer’s arms and hands. The trainer stood by the side of the tank out of the water where he or she could be seen by the dolphin. The trainer would place and move his or her arms in different positions as in a sort of semaphore signal system. To avoid the unconscious
giving of helpful cues to the dolphins, the trainers wore opaque goggles so that the dolphins could not see their eyes. The two dolphins learned to carry out correctly a number of commands in the water. The commands consisted of two-, three-, four-, and even five-word sequences, with each command constructed on the basis of object and action words. Thus, ‘window tail touch’ is to be interpreted as ‘Touch a window with your tail’. The basic sentence structure was of the Subject– Object–Verb variety.
Of special interest are Herman’s results, which show that generally the dolphins correctly responded to what are often called ‘semantically revers- ible sentences’, i.e. sentences for which the subjects and objects cannot be interpreted by meaning alone but where the use of syntactic knowledge is required. For example, the English sentences ‘Jack pushed Tom’ and ‘Tom pushed Jack’ describe two different events, one in which Jack is doing the pushing and another in which Tom is doing the pushing. Given our lack of knowledge about Jack and Tom, we can only judge that either event is equally likely to occur.
Such equal reversibility would not be the case, however, with sentences like ‘The cat chased the mouse’ and ‘The mouse chased the cat’ since, based on our knowledge of the world, we would generally expect the cat rather than the mouse to be doing the chasing. Our expectations for certain events or situations can influence the interpretation we give to words. Similarly, an animal such as the dolphin might be able to respond appropriately to a string of words, not on the basis of their structural word order but on the dolphin’s life experience. A proper test for grammatical knowledge must take this phenomenon into account.
Herman was aware of this problem, so, as part of his research, he pre- sented the dolphins with commands involving semantically reversible structures. He gave them, for example, both ‘pipe hoop fetch’ (Take the hoop to the pipe) and ‘hoop pipe fetch’ (Take the pipe to the hoop). Since the dolphins generally responded appropriately to both commands, Herman was able to conclude with some certainty that the dolphins had acquired a syntactic structure that involved relational and prepositional functions. Because word order in these commands indicates different semantic or mean- ing relationships, it is reasonable to claim that the dolphins had acquired such syntactic relational notions as direct object and indirect object.
The dolphins, Herman emphasizes, can also respond to novel sentences on the basis of understanding words and their relations in a command structure. Once the structure and relations are learned, then all new sentences with those characteristics should be understood, providing, of course, that the meaning of component words is already known. Thus, after acquiring the notions of direct and indirect object, Akea responded correctly on her first exposure to the sentence ‘person left frisbee fetch’ (Take the left frisbee to the person). Herman is therefore able to deflect any criticism that the dolphins are merely carrying out the same sort of fixed stimulus–response
type of shaped behaviour that dolphins and whales in marine parks are trained to do. He correctly points out that it could not be simple stimulus– response shaped behaviour because the dolphins respond appropriately to specific commands that they have never received before.
In later research, Herman introduced the dolphins to various notions such as Question (Herman and Forestell, 1985; Herman et al., 1993; Holder et al., 1993). A key aspect of this research is the dolphin’s ability to report on the absence of objects. In the wild, animals typically signal the presence of food or danger but probably seldom deal with topics relating to absence. Akea correctly responded to question forms such as ‘hoop Question’ (‘Is there a hoop?’) and ‘frisbee Question’ (‘Is there a frisbee?’) after searching the tank for these items. Moreover, when the question form was contrasted with the imperative (command) form, the dolphin would give the correct answer. Thus, given the question ‘hoop Question’, Akea would correctly press the ‘yes’ or ‘no’ paddles in the tank.
Herman’s research is one of the most scientific and methodologically reliable on the learning of language by animals. It remains for further re- search to demonstrate whether dolphins would be able to express in production what they have already learned in terms of language comprehension. Devising such an appropriate means of production, however, is not easy. Perhaps Herman should return to his earlier work where he trained a dolphin to mimic computer-generated sounds.