[Originally published 8.29.09]
Heaven preserve me, dolphins obey the same rules which exist in human language! They swim in accordance with linguistic rules! This is what numerous science websites, television channels and newspapers tell us following a recent study which asserted that the body motions of dolphins as they swim close to the surface obey the Law of Brevity, so that the most common actions are also the shortest. The article in question is Efficient coding in dolphin surface behavioral patterns by Ramon Ferrer i Cancho and David Lusseau and the researchers claim that this is the first case where the Law of Brevity has been documented in a non-human species, but is there really any connection to language here? I do not have a definitive answer, but there are three things I would like to talk about: the article itself, Zipf's Law to which the researchers allude and reports on science in the media.
The Language of Dolphins
Dolphins communicate using what sounds like various whistles and clicks, so the main part of their language (or rather, system of communication) is indeed oral. In a previous study, Lusseau (2006) showed how they can use different gestures in order to relay information in certain situations. These gestures are carried out when a dolphin surfaces and performes one of about 30 different motions in various degrees of complexity (see the article for a full breakdown). Lusseau mapped the gestures to the dolphins' behavior and reached the conclusion that a direct link exists between the two. But why would they communicate using gestures when they can simply "speak" to one another? Lusseau puts forward a hypothesis:
In the case of dolphins, vocalisations can be detected over many kilometres (Janik, 2000). It would therefore be advantageous to be able to communicate motivation over a short-range to avoid eavesdropping by conspecifics and therefore minimise scramble competition (Dawson, 1991).
Since the competition for food can be fierce (Lussea describes dolphins attacking dolphins from another group and even dolphins taking food from a mother and its offspring), it may be that dolphins prefer "whispering" to one another.
What does this have to do with human language? The researchers talk about what they call the law of brevity. i Cancho said in an interview:
The results show that the simple and efficient behaviour strategies of dolphins are similar to those used by humans with words, and are the same as those used, for example, when we reduce the size of a photographic or video image in order to save space.
In the lower graph, the length of the gesture is plotted against the frequency in which it is carried out. In the top graph, the length of an English word is plotted against the frequency in which it appears in Oscar Wilde's "The Picture of Dorian Gray". The researchers claim that just as a language uses the law of brevity – and I don't know anyone apart from them who calls it the law of brevity – so do the dolphins. Now, assume that's completely true. Here is one problem: 30 types of movement are not words, since the dolphins communicate vocally. Their whistles are our words; their gestures might be… our gestures, perhaps. I did not delve deeply into research on pantomime and gestures, but I get the impression that most of our gestures are relatively short. For example, when we greet someone, we shake their hand. If they are a close friend, we might even add a hug, so that if we checked our behavior over a given stretch of time, we would see that whatever is frequenter is shorter and whatever is less frequentt is also less complex. Does this mean that human gestures obey natural language rules? I doubt it. This is probably something much more basic. It does not mean that the scientists are wrong, but I think a more apt comparison might be to human gestures, and maybe-maybe even to a sign language, where entropy is very interesting. So far for a psyhcolinguistics interpretation of the results, but I also take issue with the aforementioned law of brevity.
This law of brevity is an implication of a fascinating phenomenon, which was formalized as Zipf's Law by the linguist George Kingsley Zipf. This law is very popular among computational linguist because it allows them to estimate frequencies of words in a text, like in the graph of Dorian Gray. The idea is this: if we rank all the words in a text collection by the amount of their appearances, we will find that there is a logarithmic ratio between them. In other words, if we have a hundred words, the word in the 1st place will appear twice as much as the word in the 2nd place and fifty times as much as the word in the 50th place. I am butchering the definition a bit here for simplicity's sake, but what you get is a very nice graph like the one we have above.
In an attempt to make sense of this phenomenon, Zipf claimed it was an embodiment of the Principle of Least Effort: when two people are talking, the speaker will try to use as few words as he can in order to transfer as much information in as little effort as possible. The listener will want the speaker to use as many different words as possible so that the differences would be clear to him as a listener, rather than having to decipher ambiguities. What is really fascinating is that Zipf's Law works on any text corpus, but not only on texts: also on frequency distributions the relation between which and language is as faint as the relation between a dolphin and George Costanza.
- Zipf's Law holds for the sizes of cities (although attempts have been made to disprove this): one could observe the sizes of cities in a country instead of the frequency of a word in a text and the results would be similar.
- My favorite paper on the subject shows that Zipf's Law for cities holds in Denmark. The researcher did not stop there and showed that Zipf's Law holds for the size of Danish companies as well! As a reminder, there is no real connection between language and Danish cities or companies.
- Another study showed the validity of Zipf's Law for additional cities in the USA and in Europe.
- The beauty of this distribution lies in how natural it is. Internet service providers use Zipf's Law to save frequently accessed webpages in their cache: the msot popular wepages are cached in servers close to the user so that they could be accessed more easily. There may well be minimal effort involved, but not in the sense Zipf had talked about.
So Zipf's Law is many a wondrous thing, but what does this have to do with the Law of Brevity? We can definitely notice that the most common words are also relatively short, and the floor is open for our readers to comment and tell us what they know about this topic. Either way, the fact that something, anything, obeys Zipf's Law should come as no surprise.
The science media gulped down the story about the connection between the dolphins and natural language, as is often the case. It would be hard to argue that the media are responsible for evaluating the scientific rigor of a study, and in defense of the journalists who reported on the study, it is a solid study with interesting results and not some sham. And yet, as readers we should always be vigil, since the media usually look for the sensational when reporting on science. But this is another problem entirely, which reflects nicely in the words of YNET's Efrat Weiss about a libelous report she published: it is not her job to question the authority of the IDF Spokesperson but to relay its words as is. I guess what's good for the political goose is good for the scientific gander.
I have expressed my reservations regarding the claims a number of researchers have made about the uniqueness of human language, with a few suggestions for further studies which may help prove or disprove their claims. Furthermore, I hope I managed to convey the magic of Zipf's Law through the pages of this blog, a law relevant to the frequency of words in a language, to the sizes of cities and companies and to loading internet pages. We concluded that we should be wary of science in the media, even though in this case they are not to blame. Now I feel like going swimming with dolphins in Eilat.