Why are humans smart? Language and LEGOs

posted by John

In her absolutely awesome paper “What Makes Us Smart? Core knowledge and natural language,” Elizabeth Spelke writes

When we compare the cognitive achievements of humans to those of nonhuman primates we see striking differences. All animals have to find and recognize food…but only humans develop the art and science of cooking. Many juvenile animals engage in play fighting, but only humans organize their competitive play into structured games with elaborate rules. All animals need to understand something about the behavior of the material world to avoid falling off cliffs…but only humans systematize their knowledge as science and extend it to…entities that are too far away or too small to perceive or act upon. (Elizabeth Spelke, “What Makes Us Smart? Core knowledge and natural language.” In Language in Mind. Gentner and Goldin-Meadow (eds.). 2003.)

So, Spelke asks, “What is it about human cognition that makes us capable of these feats?”

The answer to this question is a complicated one, even if you already know I’m going to say it is language. Why is it complicated? Because it’s not just language itself, but the ability, associated with language, to combine otherwise separate “core knowledge” systems. Whereas lots of animals have our same basic cognitive senses of spatial relations, object mechanics, number sense, geometric sense, and navigation, humans (once they develop language) are uniquely able to combine them and make them work in conjunction.

How do we know this? Basically, it has been demonstrated that both infant humans and many other animals have extremely similar core knowledge systems. Babies and monkeys, for example, have essentially the same ability to understand how objects move and interact, whether one group of objects is larger than another, and how basic geometry allows you to walk a room in specific, novel paths.

Each of these tasks represents a separate “core knowledge” system (you could also call them ‘modules’). Crucially, these modules in both babies and other animals are isolatedencapsulated, and unable to interface(representations from one are incomprehensible to the other).

Rats and babies—all that (cognitively) different?

To understand in what way these modules are isolated, let’s look at just one example (simplified slightly for reasons of space): Say you put a rat in a rectangular room and show him that a bit of food is located in the northeastern corner. You then disorient the rat (cruel, I know), and set him loose. Immediately, with no trouble, he will go to the northeastern corner and find the food. The rat has the cognitive ability to search using some sort of ‘directional’ or “geocentric” sense.

Similarly, if you then put a little chair in the room, show the rat that there is some food on the chair, disorient it, then set it free, it goes directly to the chair and finds the food. The rat can also do navigation by landmark.

These are two separate systems of spatial relationships and navigation: navigation by direction and by landmark. Crucially, then, if you put a piece of food northeast of the chair, the rat will search at random somewhere near the chair. This is evidence that he cannot navigate using both “northeast” and “the chair.” Combining the two systems—each of which works fine on its own—leads to problems.

Infants have the exact same problem: when directed to find something at a chair, it’s easy. When directed to find something in the northeastern part of a room, it’s fine. But northeast of the chair doesn’t work. Again, the separate modules are not able to interface effectively with each other.

Adults, of course, have no trouble going northeast of the chair. They have an ability to combine and communicate between these two cognitive systems that infants and other animals do not. The emergence of these combinatorial abilities is directly associated with the development of language. Once you can talk, you can do things like this too. How intelligent of us!

The LEGO Analogy

There’s a really nice way to think about how this whole business might work: Consider each individual module as a LEGO, but without the little raised dots on top. Each does it’s own thing pretty well—and maybe you can make a basic stack of them to do slightly complex things. But once you try anything more than the most basic of interactions between modules (LEGO blocks), your structure collapses. So when you try to combine navigational capacities to go to the left of the chair, things get confusing.

Language, then, is the little raised dots on top of the LEGO (and I guess the little holes they fit into). Once you have those, everything changes. Structures unimaginably complex from the point of view of bump-less Lego blocks now become possible. We go from a basic stack of unconnected blocks to things like a full-on LEGO arena.

Now maybe we’re not that smart—not yet at least—but that’s the basic idea. The reason that humans are smart is precisely because we have language on our side. The language capacity, Spelke and others have suggested, allows the most basic building blocks of cognitive ability to communicate and interact. So, like LEGOs with connectors, we can now build structures of near infinite complexity (remember The girl the cake the baker the owner fired baked hit screamed) and combine the faculties that previously could only work alone.

Other linguists, like Noam Chomsky or my former professor Cedric Boeckx, have taken this even further. They have theorized it’s not language, per se, that allows for communication between modules, but rather some other relatively small, yet crucial, cognitive development. Part of the core reasoning behind this is evidence that advanced cognitive abilities, like language and culture (and also the sorts of actions discussed above), developed remarkably fast by evolutionary standards. The first evidence of language goes back only some 30,000 years! Because of the relative speed with which language evolved, it’s been supposed that the critical upgrade was actually only a tiny little change, albeit with massive consequences.

Well, what if that change was, very simply, the ability to take all of the separate human cognitive faculties and allow them to work together? What if the only change was the development of a cognitive ‘connector’? We would then have the ability to take discrete modules and concepts and place them in communication with each other; the ability to build more complex structures using the most basic of building blocks. This would not only explain how our separate core knowledge systems could start to be combined, but also how we came to put words together into syntactic structures.

This theory has been influential in the linguistics world (though it’s not without its detractors). It makes some sense, too. Not only would the combination of northeast and chair be possible, we could also create structures made up of concepts based in the real world.  We could take concepts (eventually words) that previously existed as individual, non-interfacing ideas (animal, food, run), and put them together into complex thought patterns and, eventually, sentences (There is an animal that we could eat, so let’s run after it). What were previously non-connecting LEGO blocks can now be combined in majorly complex ways.

Once this ‘connector’ mechanism is sufficiently developed in human infants, they, like adults, can combine cognitive modules and, importantly, combine concepts into sentences.

As far-fetched as this might sound, it’s actually not so different from the LEGO example. You had all the blocks before, and nothing changed but the addition of connectors. That’s the only difference between the technologies, and yet it has huge consequences.

Our minds work in complex and fascinating ways, and of course there’s no way we can yet know for sure this idea is correct. But isn’t it exciting that there could be so simple and elegant an answer for why humans are smart? And you can’t deny that we are—we did, after all, invent the LEGO.