The songs of Bengalese finches tell of the origins of language

Kazuo Okanoya
Team Leader
Laboratory for Biolinguistics
RIKEN Brain Science Institute
Humans are the only animal that has language. How did human language emerge? The ultimate goal of the Laboratory for Biolinguistics in the Mind and Intelligence Research Core at RIKEN’s Brain Science Institute is to present a scenario that can reasonably explain the origins of human language from a biological viewpoint. To this end, the team takes the approach of comparing the communication behavior of a variety of animals with human language. The subjects of their research are the Bengalese finch (domesticated songbird), the degu and naked mole rat (murine rodents), and humans. “The biological origins of language are being comprehensively explored by applying findings in animals to humans, and by applying findings in humans to animals. That’s our unique method of research,” says Kazuo Okanoya, team leader at the Laboratory for Biolinguistics. Can Okanoya’s team contribute to discovering the origins of language?

The four abilities on which language is based: phonation learning, syntax, meaning and community

When entering the Laboratory for Biolinguistics, you are greeted with a number of doors to your left. From behind one, you can hear the voices of songbirds trilling ‘pi-pit’ — the sound of Bengalese finches. Standing in front of another, you can hear the feeble squeaks ‘kyut-kyut’ of degus, a murine rodent native to the Andes in South America. Beyond another door all is silent, and access is limited to a few qualified researchers to prevent transfering bacterial infections to the animals housed there: naked mole rats, which live in communities in tunnels beneath the deserts of East Africa.

The Laboratory for Biolinguistics was founded to research the origins of human language. According to Okanoya’s definition, “Language is a system in which meaningful words are rearranged according to a fixed rule to create new meanings, and it is only humans who can use language.” Then, why does he use Bengalese finches, degus and naked mole rats as subjects for his research? “If we want to discover the origins of language, we should study animals that do not have language. Language emerged suddenly in humans, but the basic abilities needed to establish language are also found in other animals. I suppose that humans might have been able to acquire language by combining the basic abilities in animals.” Examining the human brain may lead to a better understanding of how humans learn and use language. However, the origins of human language will never be discovered by exhaustive investigation of the brain of a human who has already acquired language.

Okanoya explains the basic abilities needed for language. “First, the ability to learn new vocal tones and phonate them; second, the ability to arrange the tones according to a rule; third, the ability to match the tones and meanings; and fourth, the ability to use the voice appropriately in a population. These four abilities are essential for the acquisition of language. Let’s abbreviate them as vocal learning, syntax, meaning and society, respectively,” says Okanoya. In his research, vocal learning and syntax are investigated using Bengalese finches; investigation of meaning is carried out with degus, and investigation of society with naked mole rats. He is working to elucidate the origins of human language by applying what he finds from animal studies to humans.

Bengalese finch songs involve syntax

“Have you ever heard Bengalese finches singing?” asked Okanoya. Bengalese finches, of the order Passeriformes, are familiar as pets. Songbirds such as the finches have two types of voices: calls and songs. Calls comprise short sounds, such as ‘pi-pit’ and ‘gar-gar’, which the bird makes when it is vigilant or when the chick is asking for food. Songs are mainly sung by the male for courtship or to mark territory, as exemplified by the ‘ho-hot-ket-kyot’ of the Japanese bush warbler, and the songs sometimes continue for several minutes. “Their songs resemble human songs or music. I believe that extensive investigation of the songs of songbirds will contribute to discovering the origins of human language.”

Birdsongs are analyzed using sonograms generated by a frequency analyzer, which allows the vocal signals to be visualized. When letters are assigned to ‘song elements’ with similar shapes, the songs of the Bengalese finch are found to comprise about eight different elements. “Bengalese finches sing songs by repeating a set of several of the song elements, sometimes with a change in the sequence.” For example, a song begins with ‘ab’ and this is followed by ‘cde’. Then ‘cde’ is followed by ‘ab’ in some cases and by ‘fg’ and ‘ab’ in other cases. However, ‘ab’ is always followed by ‘cde’. Each set of song elements is called a ‘chunk’, and a stream of the entire song is called a ‘phrase’. “Each chunk corresponds to a word. Bengalese finches sing songs by rearranging chunks according to a fixed rule. Hence they have syntax, just as we do,” says Okanoya.

There are many features in common between the process of a bird learning to sing and the acquisition of human language. Because of this, birdsongs have been widely investigated, mainly in the UK and the US, as a behavioral and neurological model for understanding human language since the 1960s. Okanoya’s discovery in Japan that Bengalese finch songs have syntax attracted international attention, although Okanoya is the first to admit it was a product of a chance decision.

After studying animal psychology at university in Japan, Okanoya went to the US where he earned a doctoral degree from the University of Maryland, which has a laboratory for avian auditory sensation. In 1989, when he returned to Japan and started a new research project, he faced a challenge: “The zebra finch is commonly used in birdsong research. However, if we do the same thing as the laboratories in the US and Europe, where project scales and budgets are much larger than in Japan, we will never have the chance to get ahead.” Then Okanoya noticed the Bengalese finch, which had not been used in birdsong research. “I chose it for a simple reason: it is easy to breed and propagate, but examination revealed that Bengalese finches sing more complex songs than zebra finches, and hence are better suited for research into the origins of language.” zebra finch songs are quite simple, with repeats of the song elements (e.g. ‘abcdefghabcdefgh’). If Okanoya had not chosen the Bengalese finch, he would not have discovered their song syntax.

“The syntax of Bengalese finch songs is known as finite-state syntax, and it is quite a complex rule for birdsong, but simpler than the syntax of human language. Some people think that even if we study birdsong syntax, it will not enable us to understand the syntax of human language. However, there must have been a simple stage with finite-state syntax before the emergence of more complex syntax in human language. I think that by exploring the mechanism and development of the syntax in Bengalese finch songs, we can get a picture of the beginnings of syntax in human language.”

Female birds prefer complex songs

The Bengalese finch has been domesticated from the white-rumped munia, a wild songbird indigenous to Southeast Asia, imported by a feudal lord from Nagasaki in 1762. The songs of the white-rumped munia, the original species, have about eight elements, similar to the Bengalese finch, but lack any syntax, with only simple repeats of certain song elements. During the 250 years of their domestication, Bengalese finches have therefore acquired syntax in their songs.

In Bengalese finches, song elements, chunks and phrases differ slightly among individuals. “Females seem to prefer males that sing more complex songs,” says Okanoya. An experimental study showed that on hearing complex songs, females were more active in nesting and produced larger numbers of eggs than on hearing simple songs.

Okanoya thinks that the reason why the Bengalese finch became a singer of complex songs was the result of sexual selection. “It is likely that female white-rumped munia initially preferred complex songs. However, complex songs make the bird conspicuous in the outdoors, so that singing males are more likely to be attacked by their enemies. Because of this high risk, male striated finches did not sing complex songs. Under domestication, however, there is no fear of being attacked by enemies. Food is also readily available. It is estimated that male Bengalese finches evolved to sing the complex songs preferred by the females, and the females chose males that sung more complex songs, so that the songs became increasingly complex in a short time through a process of evolution.”

Segmentation, a key to finding the origins of human language

Okanoya thinks that one key to finding the origins of human language lies in ‘segmentation’, referring to the border between sets of meaningful sounds in a stream of vocal tones where a pause is inserted. “When we listen to a foreign language that we are unfamiliar with for the first time, we do not know where the stream of utterances separates into words, but after listening repeatedly, we become able to identify the word pauses. Hence, segmentation is very important in acquiring language. With this in mind, I conducted an experiment in human adults to determine how the brain works during segmentation of sounds. We found that particularly intense brainwaves were produced at word–word pauses.” This is the first time in the world that the brain activity involved in segmentation has been visualized.

A more interesting finding was that particularly intense brainwaves are produced at word pauses in human babies, just like in adults. “Babies are unable to understand language or speak, but segmentation is an innate ability in human beings. Then what happens with Bengalese finches? We are interested in looking into this.”

Although birdcalls are innate, birds have to learn to sing. Juvenile Bengalese finches first listen to the songs being sung by their parents and the males around them. They then try singing themselves, and become increasingly good at singing after correcting their errors against the examples of the adults. Some animals learn vocalizaiton, examples of which include songbirds such as the Bengalese finch, whales and humans. Bengalese finches do not memorize the phrases sung by one male as they are, but cut and paste songs by more than one male to create their own. “In this process, they always cut and paste songs on a chunk basis. This is why Bengalese finches are capable of segmentation. We are making detailed investigations into the mechanism of this segmentation, including whether special brainwaves are produced during segmentation as in human beings, which neurons function at which timings, and which genes are expressed.”

Language originates from the songs of courtship

“The songs of songbirds convey only the one meaning, ‘I love you,’ even when the chunks are rearranged. Language can be defined as ‘something that allows its user to create a new meaning by rearranging meaningful words according to a fixed rule,’ so it is impossible to completely elucidate the origins of human language solely from birdsong. Hence, studies with degus and naked mole rats are also important. Using them enables us to deal with the issue of segmentation of situations. Provided that we can clarify the mechanisms for sound segmentation and situation segmentation, the origin of meaningful words will become evident.” Degus and naked mole rats use at least 17 types of vocal tokens in distinct ways according to the situation to maintain communication with their group.

Based on the findings in past studies of humans and other animals, Okanoya has proposed a mutual segmentation hypothesis for the origins of language. “I hypothesize that if there is a common part shared by a sound stream used in one situation and another stream used in another situation, the common part of the situations and the common part of the sounds are segmented by pauses, and they will become recognized as a corresponding pair. As a result, a word with a limited meaning is produced.” Okanoya is planning to carry out further studies to verify his hypothesis.

What then is needed to link the findings in Bengalese finches, degus and naked mole rats to the origins of human language? “The acquisition of language was an event that occurred in humans only once in the history of organisms. Anything that occurs only once can never be understood solely through biology. We must be proactive in adopting approaches in other areas, such as cultural anthropology and computer simulation.”

Finally, Okanoya offers a scenario for the origin of human language. “I think human language originated from the songs of courtship. Humans made tools, created cultures and lived in communities, so the risk of being attacked by enemies decreased dramatically. As a result, human songs of courtship become more complicated to satisfy the preference of the opposite sex, just as Bengalese finches came to sing increasingly complex songs in their cages, and this in turn led to the emergence of words with mutual segmentation — the birth of human language.”

About the Researcher:

Kazuo Okanoya

Kazuo Okanoya earned his MS and PhD degrees at the University of Maryland in the US. On returning to Japan, he spent four years as a postdoctoral fellow investigating auditory perception in songbirds. In 1994, he joined the University of Chiba as associated professor, and in 2004 he began as head of the biolinguistics laboratory at the RIKEN Brain Science Institute. Since 2008, he has concurrently served as a research director of Okanoya Emotional Information Project supported by the Exploratory Research for Advanced Technology (ERATO) program of the Japan Science and Technology Agency.