Spontaneous Symbol Acquisition and Communicative Use by Pygmy Chimpanzees (pan paniscus)
Savage-Rumbaugh et al (1986)
Savage-Rumbaugh, S., MacDonald, K., Sevcik, R. A., Hopkins, W. D. and Rubert, E. (1986). 'Spontaneous Symbol Acquisition and Communicative Use by Pygmy Chimpanzees (pan paniscus).'
Exam Board / Level:
Savage-Rumbaugh's longitudinal case study is a report on the naturally occurring use of symbols by two pygmy chimpanzees, brother and sister Kanzi and Miluka. The report is based on all data collected over a period of 17 months and demonstrates how both chimps were able to spontaneously use lexigrams (symbols) to communicate with the researchers. This was assessed on a day-to-day basis journeying around the research centre (the chimps would indicate by symbol where they wished to go and what they wanted) and also through a formal test that involved pairing spoken words photographs and lexigrams in varying combinations. The researchers compared the data to that collected on two common chimps using the same equipment and conclude that pygmy chimps (pan paniscus) are better predisposed to acquiring language than other great apes.
AbstractSavage-Rumbaugh et al's case study reports on the 'spontaneous' acquisition of 'language' in two pygmy chimpanzees, who started to use symbols to communicate with people based on observation and without any training. Both animals demonstrate the ability to connect spoken words in English to lexigrams (printed symbols), something common chimpanzees are unable to do. Most of the information collected relates to Kanzi, the older, male pygmy chimpanzee, who was able to use the symbols to create telegraphic speech. The results indicate that pygmy chimpanzees have different symbolic and auditory skills to common chimpanzees.
BackgroundPrevious research, such as that undertaken by Gardner and Gardner (1971) with Washoe, a young female chimpanzee, demonstrates that apes are able to learn to use a language (American Sign Language or other symbols as they can not use spoken language), or at least, learn to use symbols in order to bring about desirable events. It has also been shown that once they have been taught how to label objects, make requests and understand they are able to create novel utterances indicating their intentions or pointing out aspects of their surroundings (indicative referential symbol usage).
Apes learn language in a very different way to human children, as they require extensive training, where the symbols for objects and actions are presented to them and they are reinforced for producing the correct symbols. Loulis, a baby chimp adopted by Washoe, was seen to use hand gestures without training, although was not formally assessed or observed. Therefore it is impossible to know if these gestures serve any function, as chimpanzees naturally use hand gestures in the wild and in captivity.
There is also a difference between associating a symbol with a specific object and developing the understanding that the symbol can be used to represent the object. Young human children gradually learn to use words in a representational way and it may be that chimps such as Loulis have not made that transition.
Lock (1980) outlines the stages involved in the shift from associative language use to representational use. The first stage is 'ritualised games', such as naming pictures in a book, where a parent asks the question and then answers it. The child learns to answer the question but initially does not understand the symbol they are using, other than that it is linked to a specific picture. Over time they learn that the symbol (eg. the word 'ball') represents the object associated with it, and eventually can respond appropriately in the absence of the original context (eg. can go out of the room and get their ball when asked to do so).
All of the previous evidence (including Savage-Rumbaugh's own studies) shows that apes are unable to naturally transition from associative to representational use of language without extensive training, although once they do, their linguistic development follows the same pattern as described by Lock in human children.
Previous studies also fail to demonstrate whether apes can understand spoken English words. Although Gardner and Gardner suggest Washoe had this ability, there was no evidence to support this. Other apes who have been reared in human homes have also failed to show comprehension of spoken language, even though they have been exposed to it (human children learn in this situation).
The only exception is Koko, a gorilla raised in captivity and exposed to speech and gestures from an early age. Koko was assessed using the Assessment of Children's Language Comprehension Test (Patterson and Linden, 1981), which presents the subject with four drawings and a sentence, with questions becoming increasingly more difficult so that younger children tend to do worse as the test progresses and surpasses their current level of language comprehension. Older children tend to perform well throughout the test. Koko on the other hand got about half of the answers right throughout the test.
The study conducted by Savage-Rumbaugh et al is a longitudinal case study of language acquisition without training in two pygmy chimpanzees (pan paniscus): Kanzi and Mulika (Kanzi's younger sister). Both animals developed the ability to understand language without any systematic training.
The study is not an experiment, but is a description of events that occurred naturally. Previous evidence indicates that pygmy chimpanzees (one of the four species of great apes) have different social arrangements to the other species, with males taking part in infant care, frequent food sharing, even between adults, and strong male-female ties.
Eye contact, gestures and vocalisations are also more frequent in this species. Cognitive studies, such as that conducted by Yerkes (Yerkes and Learned, 1925), also suggest that pygmy chimpanzees are brighter than other great apes. All of this suggests that pygmy chimpanzees are better predisposed toward language acquisition.
- 2 common chimpanzees (pan troglodytes) for comparison, raised in a language-using environment by people between 1975 and 1983 - Sherman and Austin
- 2 pygmy chimpanzees (pan paniscus) - Kanzi and Mulika, raised in a similar environment to Sherman and Austin, but with access to their mother.
The main subject was Kanzi, aged four at time of publication, raised at the Yerkes Regional Primate Research Center from the age of six months. Both Kanzi and Mulika spend several hours a day with their mother, but choose to spend a considerable amount of time with humans.
Austin and Sherman were 1.5 and 2.5 years old and were removed from their mothers before going to the research centre - aged 9 and 10 at time of publication.
MethodologyThe medium of communication was a system of geometric symbols that brighten when touched - an action that is treated as being the same as uttering a word.
Kanzi's keyboards were connected to a speech synthesiser which speaks the words when the symbols are touched. Lana - the first chimpanzee to use this system, used a special language called 'Yerkish' and the symbols were colour coded according to function, but both of these features were dropped for later subjects.
Technological development means that Lana, Sherman and Austin's keyboards were all connected to computers and therefore were inside, whereas Kanzi had battery operated keyboards and a portable voice synthesiser that could be taken outdoors, although most of the time Kanzi and Mulika used a pointing board containing photographs of the lexicons. These boards did not work with Sherman, Austin or Lana.
Spoken language accompanied the use of the symbols, along with around 100 American Sign Language signs (ASL - none of the researchers were fluent in ASL). All modes of communication were used and accepted.
Kanzi was exposed to his mother's (Matata) training using the lexigrams, as it was not possible to separate him whilst she was training. Most of the time he showed no interest in the symbols and often interrupted her training. At one and a half years of age he started to show an interest in the symbols, lighting symbols and running towards the vending device, indicating he had learnt that touching the symbols made the device dispense food. He began to use the chase lexigram and also a clapping gesture to initiate chase games.
At two and a half Kanzi was separated from his mother whilst she went to breeding colony, returning 4 months later, pregnant with Mulika, who was born nine months later. On her return, Kanzi chose to spend more time with humans. Mulika was separated from Matata at four months of age for treatment of an eye infection and after this also chose to spend more time with humans. Mulika did not observe Matata's training, but she did observe Kanzi's.
Sherman and Austin
- Introduced to lexigrams via training, but did learn some in an observational setting.
- Did not use keyboard outside of lab.
- No speech synthesiser as tests showed no speech recognition, but tonal cues used.
Kanzi and Mulika
- Introduced to / used lexigrams in an observational setting.
- Portable keyboards and laminated pointing boards outdoors.
- All use of symbols accompanied by speech, either synthesised or spoken.
In warmer months food was dispensed in 17 locations around the research centre’s 55 acres of forest and no food was available in the lab. Most of the day was spent outdoors travelling from place to place to obtain food. The researchers initially left photographs of the different foods at each location and Kanzi pointed at the food to indicate where he wanted to go next, often taking the photograph with him, pointing at it and vocalising along the way.
Over time Kanzi learnt the locations of all the foods and could lead the researchers to each location. Mulika started to travel with Kanzi from 6 months, Kanzi often taking turns to carry her. By her second summer she was old enough to share the foods and started to initiate travel herself.
If Kanzi did not want to eat the food at a given location, he was asked if he wanted to take it with him in a backpack. If this question was ignored the food was left where it was. Alternatively Kanzi would indicate by touching the backpack and vocalising that he wanted to take the food with him.
During the course of the day foods would accumulate in the backpack, so a symbol could relate to the food or the location where the food was. Kanzi was asked to clarify and would do so by touching the backpack or indicating in the direction he wished to travel.
Whilst Kanzi and Mulika were indoors they were asked to help with tasks such as changing sheets and doing laundry but often tried to initiate play and would spontaneously try to assist with tasks such as washing dishes, spraying the hose and moving items. They played with clay, bubbles and dolls and had special video tapes of people they knew doing things that were interesting to them. Sometimes they were entertained by people dressed in animal costumes. Kanzi usually asked to watch TV in the evening and the researchers created a number of video tapes of interest to chimpanzees, with lexigram overlays.
Measurement and Analysis
All use of communication was recorded for Kanzi between the age of 30 and 47 months, and for Mulika between 11 and 21 months.
Indoors all data was collected via the computer. Outdoors the researchers recorded manually and then transferred the data to the computer at the end of the day.
Utterances were recorded in three categories:
- spontaneous (those initiated by Kanzi or Mulika without prompting);
- imitated (those containing anything from the researcher's previous utterance);
- structured (those in response to a request or question).
Utterances were further coded as requests, responses or statements.
The criterion used for when vocabulary had been acquired was that:
- the use had to be spontaneous on 9 of 10 occasions;
- had to be verified on 9 of 10 additional occasions (behavioural concordance).
Examples of behavioural concordance: if Kanzi asked to go to the treehouse, acquisition was only recorded if Kanzi then led the researcher to the treehouse; if Kanzi asked for a specific fruit then he either had to go to that location or select that fruit from the backpack. If the only food in the backpack was the one that was requested, then this was not recorded.
If at any time an acquired symbol dropped below the .90 criterion it could be removed from the recorded vocabulary.
Kanzi was tested informally in everyday situations, for example being asked to name the object he was playing with or to pick up objects in a specific order. These occurrences were based on already acquired vocabulary.
Both Kanzi and Mulika were tested formally at the end of the 17 months covered by the report, when Kanzi was aged between 46 and 47 months. Mulika was tested between 18 and 21 months.
Prior to this, Kanzi had been tested:
- At 32 months on receptive skills;
- At 35 months on naming skills;
- At 45 months on naming and receptive skills.
Formal Test Procedure:
- 3 or 4 items were presented in every trial;
- the test items and alternatives were randomly determined;
- the items were not repeated on consecutive trials;
- each item served as a test item and an alternative;
- all photographs were the same size, with different photographs of the same objects used;
- all lexigrams were the same size and were replicas of those on the keyboards;
- alternatives arranged behind a blind by a second researcher, so the first researcher could not see them and give away visual cues.
Types of Trial:
- Photograph to lexigram: select the correct lexigram from 3 alternatives;
- Spoken English to photograph: select the correct photograph from 3 alternatives;
- Spoken English to lexigram: as above but with lexigrams;
- Synthesised speech to lexigram: as above, but used to control for the possibility that voice intonation influenced the response - only Kanzi was tested on this.
Sherman and Austin were not tested on the spoken trials as they did not show any recognition and did not like these trials. They were also rewarded with food for correct answers to facilitate co-operation.
Four months after the foods were first placed in the forest a blind test was conducted with a researcher who did not know the locations of the foods, the trails through the forest, or how to return to the lab. 6-10 alternatives were presented to Kanzi at each location and he successfully travelled from one location to the next.
As well as the controls identified above, the formal tests were used to control for contextual cues. The order of presentation and the location of stimuli were controlled to reduce experimenter effects on his response.
4.5 hours of interaction was video-taped without the researcher's knowledge that this would be used to check reliability. An independent observer recorded from the video tape and there was 100% agreement on the symbols used by Kanzi and whether these were correct. There was disagreement on the usage of one (whether it was spontaneous or structured) and 9 utterances were noted from the tape that were not seen 'real time', but in all cases Kanzi was able to get the researcher's attention by repeating himself.
ResultsBetween the ages of 6 and 16 months, Kanzi and Mulika started to spontaneously use symbols to communicate, for example: Kanzi would throw a nut to a person and if they did not understand, would put a small rock on top of the nut to indicate that he wanted them to crack it open.
Mulika was also able to use this strategy to request that a balloon be inflated. Matata was less able to use symbols in this way, but occasionally placed a hose in a person's hand and indicated at Sherman and Austin if they had displayed at her.
Sherman and Austin were less able to elaborate on their communications. For example, Sherman would scream and indicate he was frightened but could not specify what had frightened him, whereas Kanzi would go and get help and clearly point to the feared object.
Kanzi searched the keyboard for the appropriate symbol and would use it to ask for particular foods even if he didn't want to eat them. If he requested a particular food and was brought three different foods he would select the one he wanted and ignore the rest.
Kanzi requested an apple, took a few bites of it, dropped it then indicated chase and started a running around the room. He then picked up the apple and indicated 'apple chase' and ran with the apple tucked between his leg and abdomen.
Kanzi touched 'ball' then searched for the ball, which had rolled under the keyboard. He grabbed it and started slapping it vigorously.
Mulika began using symbols at 12 months and initially used the symbol 'milk' for everything, pairing this with gestures indicating what she wanted. At 14 months she requested milk and was given it, but refused to drink it. She searched the keyboard and found the symbol for 'surprise', to which she was given an unusual food from the refrigerator. After this she started to use other symbols and would only use milk if she was confused.
Kanzi and Mulika initially used symbols in specific routines for some time before understanding their referential function. For example, Kanzi was told of strawberries when he was picking and eating mushrooms on the trail and was taken to the strawberries. After this he initially only referred to strawberries at the mushrooms location, but later used the symbol at the strawberries site and would respond by grabbing the strawberries if someone said 'hide strawberries'.
Mulika's use of symbols started earlier than Kanzi's but her progress was slower, perhaps because Kanzi had learnt a lot more when he was with his mother than was evident in the data. Also Mulika was not required to demonstrate concordance as often as Kanzi and the formal data for Mulika was relied on more than for Kanzi. Mulika was able to demonstrate behavioural concordance on 85 of 86 occasions when tested.
As with human children, Kanzi's recognition of symbols preceded his production of them for 63% of the words in his vocabulary. More than 80% of Kanzi's utterances were spontaneous, with only 11% being imitated or prompted. Kanzi may have relied more on imitation if he had begun to use the symbols earlier - his mother's keyboard consisted of only 8 symbols and the keyboard was expanded along with Kanzi's abilities, to 256 symbols at time of publication. Mulika experienced the complex keyboard from the outset.
Kanzi started to combine symbols within the first month of spontaneous use of the keyboard and within the 17 months produced 2,540 combinations, 764 of which were unique. This was only 6% of his total utterances and all of his three word utterances were to instruct someone do something to someone else (eg Person tickle Kanzi). 36% of the time he was not the beneficiary of the action and he would also ask for one of the researchers to perform the requested action to another of the researchers - this was initiated by Kanzi and not imitated.
Kanzi and Mulika tended to imitate most when learning new words and it could be concluded that this is a strategy used by all language learners, where the model shows them what to say first.
Kanzi and Mulika also used a significant proportion of truly spontaneous symbols, compared to Sherman and Austin, whose symbol use was categorised as spontaneous but often followed an action made by the observer to elicit the response.
In the tests, Sherman and Austin demonstrated frustration (scratching all over) in the language tests, tried to leave and vocalised in a way they had learnt since interacting with humans, not a sound that is naturally part of their repertoire. Their symbol recognition also dropped to chance in these trials.
In contrast, Kanzi was able to correctly identify symbols in the spoken and synthesised language conditions, although was less accurate in the synthesised trials. However, this was related to specific objects and people also had the same difficulties understanding the words on these occasions.
Blind Trail Test
On the blind trial, where Kanzi went on the forest trail with a naive experimenter, he followed the most direct routes to all locations that he had previously indicated he intended to go to, pointing at the photographs or symbols along the way.
Only on one occasion did he not take the most direct route, and this was to visit a place where he was not normally allowed to go. The researcher asked him to lead them to two locations they had not visited and Kanzi obliged, again, by the most direct route.
Kanzi used one symbol to generalise to a group of similar objects (eg. tomato to refer to all small red fruits), sometimes grouping them in ways that did not occur to the researchers. Kanzi also used the symbols for foods located at various places in the forest to indicate the intention to travel there. He could comprehend the names for the locations, as when asked to lead the researchers there he would do so, but would most usually use the food name when asked when he wanted to go.
Kanzi was frequently observed to take the keyboard and go off on his own. He was seen to touch the symbol for 'pine needle', then go and collect pine needles, or collect small pebbles and put them on top of the 'rock' lexigram. When the researchers attempted to interact with him in these situations he abruptly ended the interaction.
He would also use the keyboard to express his intention to behave differently: for example if he did not want to be where he was. On one occasion when he had been confined for repeatedly eating wild mushrooms he touched 'no bite mushrooms' - when he was asked if he would be good outside, he responded with positive vocalisations and subsequently behaved well when he was taken outside.
DiscussionIn previous studies, selected evidence has been used to demonstrate the ability to use symbols in appropriate contexts, when there may also have been considerable evidence demonstrating inappropriate use. Savage-Rumbaugh et al therefore present all of the data available and highlight that this was compiled from daily notes over an extended period of time.
As Matata did not spontaneously use language this suggests that there is a critical age for learning language in pygmy chimpanzees. However other variables in rearing and environment may also be a factor, and different species are being reared in the same environment for future research.
Although the sample was very small, the comparisons to other primates are useful in demonstrating the language abilities of pygmy chimpanzees. Common chimpanzees showed no understanding of spoken English and were more likely to generalise and use symbols interchangeably within categories - this was also observed with Washoe (Gardner and Gardner). In contrast, Kanzi and Mulika could discriminate between items within a category (eg juice and coke).
Further comparisons show that common chimpanzees were unable to combine phrases to suggest A act on B when neither A nor B referred to themselves. Indeed, the beneficiary was always the chimp in Sherman and Austin's combined utterances of this type. Savage-Rumbaugh et al suggest that this ability to symbolise complex interactions between others may be the basis for the occurrence of syntax.