It's a Jungle in There: How Competition and Cooperation in the Brain Shape the Mind (32 page)

My expression of humility may seem at odds with the boldness of the pronouncement given earlier in this volume, that I have an idea for a global theory of cognition—or rather that I think such an idea has been available since the mid-1800s. There’s no contradiction between hopefulness and humility, however. The main impetus for the theory I’ve sketched is the acknowledgment that the field of cognitive psychology (or cognitive science more generally) is in need of an all-embracing principle. As I said in the Preface, cognitive phenomena are generally presented as a rag-tag collection of curiosities—what might be found at a psychologists’ tag sale. It takes humility to say that this is the state of affairs and that we should try to do better.

How can you and I search for a fuller account than the one given here? In the remainder of this chapter, I’ll try to point the way by addressing several issues that haven’t been addressed yet. For all of them, I’ll argue that the Darwinian approach is promising. First, I’ll discuss concepts and conceptual development. Then I’ll consider emotions, sex, and altered states of consciousness. In the section after that, I’ll comment briefly on social cognition. Consciousness and qualities of experience (qualia) will be the focus of the subsequent section. The section after that will be concerned with evolutionary psychology, an important area of contemporary research to which this
book bears similarities but also important differences. The transmission of ideas between individuals will be covered in the next section. The origin of language will come up in the section to follow. In the penultimate section, I will say more about the sort of theory we should strive for. Here I will defend the view that the sort of theory offered here, inspired by Darwin, is the kind of theory we should pursue. In the final section, I will review how the jungle view could be applied to all the topics covered here and end with an urge for a critical consideration of the general proposal offered in this book.

A topic I didn’t cover before was the mental representation of concepts. An early proposal about concepts was that they are mental entities that can be captured by rules. An
animal
, for example, is a living thing that eats, breathes, and so on. A
bird
, meanwhile, has those attributes and others that other animals don’t. Birds have feathers, for example. Because of this exclusive feature of birds vis-à-vis other animals, it’s possible to entertain the possibility that in people’s minds, the concept of
bird
is a subset of the concept of
animal
. If that’s the case, it should be possible to place the concept of bird beneath the concept of
animal
in a neat and tidy tree diagram of the sort an executive of a large corporation would approve.

If this hierarchical structure does indeed characterize the way concepts are mentally organized, less time should be needed to affirm that a bird has feathers than to affirm that a bird eats. In other words, if you read a sentence like “A bird has feathers” or “A bird eats,” you should need less time to press a “Yes” button following the first sentence than the second. The reason is that “has feathers” is more directly linked to
bird
than is “eats.”

This prediction was confirmed in an experiment designed to test it for a broad range of sentences, not just the pair given above. In general, the more levels that had to be accessed to arrive at a correct “Yes” answer, the longer people took to give that response.
³

This outcome is consistent with the hierarchical model. However, other findings called the model into question. These other results showed that concepts are not mentally organized in a neat, tidy hierarchy. Rather, they occupy a much “wilder forest” where they engage in a kind of free-for-all, clamoring for recognition.

What findings challenged the hierarchical model? Two were especially damaging. One was that verification times were not always best predicted by the number of levels that supposedly existed. For example, the time to
confirm that “Apples are eaten” was less than the time to confirm that “Apples have dark seeds.”
⁴
This outcome goes against the hierarchical model because “have dark seeds” should be stored directly with apples. By contrast, “are eaten” should not (or need not) be stored with “apples.”

The obtained result—that people are quicker to confirm the “are eaten” proposition than the “dark seeds” proposition—makes sense from the perspective of how often people say things about those concepts. You’ve probably said, “I like to eat apples” or “Apples are good to eat.” It’s less likely that you’ve said, “Apples have dark seeds.” It makes intuitive sense that if you’ve said something often, your saying it over and over helps you confirm it quickly. Yet according to the hierarchical theory, at least in unelaborated form, how often you say something shouldn’t affect the time you need to say it’s true.

The other result that’s problematic for the hierarchical model came from studies of typicality. If an example of a concept is typical (usual) or atypical (unusual), that shouldn’t affect how easily you can confirm that the example illustrates the concept, at least according to the (unelaborated) hierarchical view. However, typicality matters. If you think of examples of birds, you’re more likely to think of “robin” or “canary” than “kiwi” or “ostrich.”
⁵
Likewise, if you indicate whether a picture is an example of a category like “bird,” you’ll probably take longer to indicate that an ostrich is a bird than that a robin is a bird.
⁶
Finally, if you rate how sensible sentences are, you’d probably give a higher sensibility rating to the sentence “A robin flew down and began eating” than you would to the sentence “An eagle flew down and began eating.”
⁷
This difference would not be expected if you checked propositions hierarchically.

One way cognitive psychologists have dealt with these results is to say that concepts are mentally represented in terms of stored instances. With enough stored instances, you can identify a most typical or “prototypical” example. You can also represent concepts with abstract generalizations—blanket overviews that apply to all members of the corresponding sets. Abstract generalizations are often expressed in statements like “an even number is a number that leaves a remainder of 0 when divided by 2.”

There is no
a priori
reason why abstract generalizations and prototypes can’t both be used by people in their everyday lives. In fact, it appears that both of them are. So you may understand what an even number is in the sense that you can generate an abstract sentence like the one I just offered, but you’d be more likely to say “2” when asked to name an even number than you would be to say “438.”
⁸
Neither number is any “more even” than the other, of course.

How a concept is mentally represented can be affected by how it’s learned. A demonstration of this principle came from a study in which two groups were
taught the same concept in different ways. One group was taught the concept in a way that favored the stored-instance method. The other was taught the same concept in a way that favored the abstract-generalization method. Brain scans of the participants in the two groups, taken after they learned the concepts, showed that different areas of their brains became active depending on how they learned the concepts. For the stored-instance group, the occipital visual areas and cerebellum “lit up.” For the abstract-generalization group, the prefrontal cortex became active.
⁹

How do this result and the others just summarized bear on the jungle principle? First, the results suggest that concepts are not represented in terms of neat, tidy hierarchies, as already mentioned. Second, the results suggest that concepts are represented by teams of self-interested parties. All the team members metaphorically express their enthusiasm or relative lack of enthusiasm all the time. Their “votes” converge on the output system that ultimately interfaces to the outer world.

Where this leads, in my view, is that there are no concepts
per se
, or none in the sense that any concept can be equated with a specific neural state. What there are, instead, are neural networks that tend to become excited or inhibited given inputs they receive. Which neurons become excited or inhibited and to what degree defines the behavior that, from outside, suggests particular concepts within.

Concepts do obey at least one clear and powerful constraint: They build on each other. You can’t grasp a concept in tensor calculus without understanding the basics of arithmetic. You can’t design advanced electrical circuits if you don’t know the rudiments of electronics. It would be strange to encounter someone who understands higher concepts without appreciating the concepts’ foundations. No matter how obvious this is, it provides still more evidence for the jungle view. As I’ll try to show below, concept development doesn’t follow the course it does just because concepts are logically dependent on one another. Concepts also develop as they do because of the dynamics of competition and cooperation within the brain.

To see what I mean, consider how you might raise doubts about Darwin’s theory of evolution. You could look for data showing that some species appeared much earlier than Darwin’s theory would predict. Suppose you joined a team of paleontologists who found a regular progression of complexity among specimens unearthed in a dig. The higher the layer, the more
complex the organism, and the lower the layer, the simpler the organism. In the midst of their digging, while already deep down and encountering nothing but the most primitive organisms, the paleontologists suddenly find, amidst the rock layers—no loose gravel or sand here—the remains of an
ape
. “What’s this doing here?” they exclaim. They bring all the specimens to the lab to find out how old they are. Sure enough, the tiny invertebrates lived millions of years ago, and the deeper down the invertebrates were, the longer ago they lived. Much to everyone’s surprise, however, the proto-ape lived right in the midst of the bygone times.

Finding an ape in the midst of millions-of-years-old invertebrates would upset the Darwinian applecart, for Darwin’s theory describes
evolution
, not
revolution
. Finding ape remains in the midst of other, more primitive fossils would cast doubt on Darwin’s theory.
¹⁰

Why speak of too-soon apes? The reason is that, just as it would be surprising to discover an advanced species much too soon in evolution, it would be surprising to discover an advanced concept much too soon in learning. A superstar kid might be able to understand tensor calculus at an incredibly young age, but she would be unable to grasp tensor calculus before understanding numerators and denominators. A brilliant child engineer might be able to grasp the intricacies of integrated circuits, but she wouldn’t be able to do so before learning Ohm’s Law.
¹¹
Concepts build on each other. They’re learned in ways that make it impossible for more complex concepts to be formed before simpler ones are established. Educators know this and develop curricula accordingly. Parents know this, too, and teach their kids simple things before teaching them more complex things. Each of us knows this as well insofar as we choose to learn what we think we can handle so far.

A parallel exists, therefore, between the development of concepts and the development of species. Just as species can be placed in an evolutionary tree, so can concepts. When you were growing up, you didn’t always know what dogs were, for example. When you were little, you may have called all furry animals “doggy.” “Doggy, doggy!” you may have exclaimed while pointing at a beagle or a poodle. “Doggy, doggy!” you may have also called out while pointing at a cat or goat. You may have applied the “doggy” name to anything with four legs and fur. The doggy concept served you well at the time. Among other things, your calling too many animals “doggy” may have led your mommy or daddy to smile and nod and say things like “Yes, Nora, that’s a doggy” or “Yes, Sarah, that’s a doggy.” My wife and I said those things to our daughters, Nora and Sarah, when they were little. We encouraged them to express themselves, even if the words they were using weren’t exactly right.
Parents, in general, are more apt to reinforce cuteness than conceptual or grammatical accuracy.
¹²

Developmental psychologists have traced the evolution of young children’s concepts through a variety of techniques. One is preferential looking. Here, children are shown single pictures and then are shown two pictures at a time. One of the pictures in the pair belongs to the same category as the single picture shown earlier; the other does not. If a child appreciates what it means for a picture to be in a category, then he or she should look longer at the out-of-place picture than at the in-place picture.

This prediction has been supported so often enough that it has come to be used with a wide range of categories to discover children’s classification schemes. With this method, it has been shown that over the course of development, children’s concepts become more and more differentiated. Though young children see cats and dogs as being in the same category, as reflected in their early-age looking, older children see felines and canines as belonging to different categories.
¹³

The other method for studying conceptual development uses questions and answers. Here, kids answer questions like “Is it silly to say that a pig can be sorry?” or “Is it silly that a rock can be an hour long?” Using this method, Frank Keil of Yale University inferred children’s conceptual taxonomies. He did so by determining the descriptors children accepted for various categories.
¹⁴
If a child said a pig could be sorry and that a person could be sorry, Keil placed pigs and persons under the same “sorry” node. Continuing this exercise for children at different grades in school, Keil showed that children’s conceptual taxonomies become more and more differentiated as they advance to higher grades.