How Can You Recognize Language?

Coach signals a play. Is that language or something else?

This is a blog about the origins of speech, but what began? How can we tell it when we see it? Parents usually say their children have started talking when they have a couple of words. Linguists tend to look for some hint of grammar. Some experts look for a favorite generative procedure. So there is room for argument even before we come up with a single fact about the beginnings.

I have always taken a functional approach. When I started this blog I could not have said precisely what I was looking for, but now I know what I want. I call it language when a speaker and a listener exchange news about a topic. The closest thing in the animal kingdom to this kind of behavior is the waggle dance of the bee. At one time that would have bothered me and I might have looked for some way to fiddle with the definition, but now I don't much care. Bee dances show that cooperative societies can share knowledge. We do it. If bees do it, and if tomorrow elephants do it too, that's the way it is.

But there is a long list of "design features" that characterize human language. In the 1950s and 60s the linguist Charles Hockett worked out a list of properties that, taken as a whole, were supposedly unique to language, and the list has become one of the commonsense tests of language origins. If your theory ends up with something that includes Hockett's properties, you may be onto something; if not, not. Since Hockett's day some of these properties have been identified in animal communications. Three persist as unique to humans. They are:

• Displacement: the ability to refer to absent, even non-existent, things. Frankly, I wouldn't be surprised if ravens and elephants can do as much. Captive apes have done it; e.g., signaling they want a hug even though there is no hugging going on around them.
• Duality: the ability to produce an unbounded number of different signals while limiting oneself to a finite number of elements. The day this property turns up in some other lineage than our own will be quite a day.
• Productivity: the capacity to coin new words or phrases that are immediately understood by others. Captive apes have also shown this ability; e.g., signaling 'water bird' for swan (unless it was signaling that it saw water and, simultaneously, a bird.)

These properties all reflect a deeper peculiarity of language: it enables speakers to report on novelties of the mind, not just on what is known about the here and now.

Here's an example of language at work, full throttle: Smith Henderson's new novel Fourth of July Creek includes the discovery by a young girl "That your daddy could drown you on accident." [end of chapter one] Yeah, yeah. The official usage is "by accident," but this little girl is discovering something new and isn't up on the standard usage. She uses "on accident" instead of "on purpose" and both her meaning and its contrast is fully understood. That's the beauty of language. You learn something new; you can improvise a way of stating it.

The history of the human lineage is the story of people learning new things and jerry-rigging a way to say it. That's what unifies Hockett's design features and gives even the shared properties  a uniquely human twist:

• Arbitrariness: there is no necessary link between signal and meaning. I find that many people resist this idea, but it is essential to linguistic freedom. We can say anything because we are not limited by some prior logic that connects certain signals with certain meanings.
• Correctability: We can change our sentences as we speak. We can take back words, revise and edit them. To me, this ability has always suggested that language is a two-system tool. One system is for recalling the words we need to use, the other is for comparing what we say with what we hope to say. This property means that our thoughts need not be prisoners of language.
• Culture: Languages are passed on from generation to generation and alter as time goes by. If it were not for this feature we would not be able to build on the improvisations of generations past.
• Discreteness: Language is produced as a stream of units that are perceived separately and can be reproduced. Thus, I can say, "Four score and sixteen years ago," and you can pull sixteen from the flow and discuss it. Duality and productivity both depend on this feature.
• Interchangeability: Anyone can produce any sentence. I can say, "I'm Abraham Lincoln," even though I am not Mr. Lincoln. Many animals communicate their status, but it is rare (unknown?) for them to pretend to a title not their own. Many eusocial insects give off scents indicating a function. They cannot lie or pretend their way through the colony by making up these scents. Language does not have these limitations. The price we pay is that we are a species of liars, but also a species of dreamers, an ambitious species.
• Reflexiveness: We can use language to talk about language. I think that really the feature is broader. We can talk about any topic we wish (except for the sentence we are producing as we speak).
• Semanticity: Words mean something. On this blog that claim says words direct the audience's attention to something. Apes sometimes deliberately catch one another's attention but they do not direct attention to a topic.
• Specialization: We intentionally use language. It is not a reflex as many animal communication systems are. There are many weaknesses with this idea, but it is true that along with all the habitual and reflexive sides of language, there is a conscious, deliberate side. And it is true that important speech is intentional speech.

There are different lists of Hockett's features, but these 11 will do. They make the point pretty sharply that language—whether on purpose or on accident—has many qualities that make it easy for our species to turn thoughts into public property.

Now a new article has appeared titled, "Language Evolution: Why Hockett's Design Features are a Non-Starter." (available here). The authors are a couple of Polish linguists at Nicolaus Copernicus University, Sławomir Wacewicz and Przemysław Żywiczyński. I congratulate them on the bold directness of their title and paper. I can't quite congratulate them on the paper itself.

Their conclusion is that Hockett's properties are no guide to how language evolved. I can hardly disagree, as I have never found the properties useful in thinking about origins myself. On the other hand, so what? Hockett's list provides a pretty good test of whatever theory you do come up with.

Suppose you want to argue—as many fine investigators do—that language arose from gesture. Chimpanzees in the wild gesture all the time. What's more, we still gesture as we speak and there is no reason to assume that there was never a time when gesture was not part of speech. So we have to at least think about the possibility that the lineage had a period when it used a gesture language before we had speech.

What is the difference between gesture and language? Gestures are signals, but, as used by apes, they are not used to signal news about a topic. Apes gesture to control others: go away, give me food, etc. Some human gestures, however, do signal news. The New York Times recently published a report on a widespread system of gestures used by church ushers. The gestures can report requests, seat availability, problems and changes in status of the church service. These signals are definitely used to tell each other some news, something unusual in the animal communication business.

How do these news-telling gestures fit with Hockett's properties? Some, indeed are present. Missing, however, are the features related to discreteness. The usher's gestures do not flow from meaning to meaning. The ushers signal one piece of information. Without discreteness a gesture system cannot have duality or productivity. It also lacks reflexiveness, the ability to add improvised topics to the system.

Thus, it looks to me like any use of systematized gestures to tell one individual something is a dead end. Even if one was used, it had to be replaced by either a sign or spoken language. So, while I agree with Wacewicz and Żywiczyński that Hockett's property list is no place to begin an inquiry into language's origins, I think it still has something to offer in testing any hypothesis that emerges from the search.