It's a Jungle in There: How Competition and Cooperation in the Brain Shape the Mind (36 page)

I won’t review Dawkins’ arguments here or cover similar arguments offered by another influential author, Daniel Dennett, in his fine book,
Darwin’s Dangerous Idea
.
⁵⁹
Instead, I’ll review two recent findings that illustrate the power of this approach. Neither finding was reported in Dawkins’ book, or in Dennetts’, for both of the findings came out after those books appeared.

The first finding concerned what’s hot and what’s not. I refer not to climate but to culture. It turns out that words and phrases appearing in print rise and fall like species in the wild. Take the term “1950.” Before 1950, there weren’t many in-print mentions of “1950,” but as 1950 approached, its mention climbed. In the year of its name, “1950” was mentioned most, but afterward its mention dropped. By 1965, “1950” was mentioned about half as often as when it was mentioned the most.
⁶⁰
Thus, the half-life of the word “1950” was about 15 years.

It turns out that the half-life of year names has gotten shorter over time. This led the authors of the study in which this finding was reported to conclude that “We are forgetting our past faster with each passing year.” At the same time, the frequencies of year mentions have grown. In other words, years are mentioned more often now than they were 50 years ago or, assuredly, 1500 years ago.

What do these results imply about “inner jungleism”? They show that concepts in culture obey dynamics similar to concepts in people’s minds. The evolution of ideas, whether represented by words appearing in print or by neural firing patterns of individual brains, can be studied profitably in selectionist
terms. Being able to analyze data of seemingly different kinds in the same general terms is the hallmark of a good theory.

The last empirical finding I want to summarize in this book concerns the number of phonemes in different languages. A phoneme is a sound that differentiates words within a language. The sounds “r” and “l” are phonemes in English, so “right” and “light” are two distinct words for English listeners. But because “r” and “l” are not distinct phonemes in Japanese, “right” and “light” are not two distinct words for Japanese listeners. English listeners are likewise unable to hear some phonemic differences in other languages.

Some languages have few phonemes, but other languages have many. It turns out that the number of phonemes in a language is related to the number of speakers of that language. The smaller the number of speakers, the smaller the number of phonemes. This means that pockets of people in some locales may use phonemes that other groups of people in other places don’t. In general, languages spoken in regions near one another tend to have more phonemes in common than languages spoken in regions more remote.

Given these facts, it should be possible to develop a tree of life for languages, similar to Darwin’s tree of life for species. It should be possible, in other words, to draw a tree in which the languages of the world spring from one another, with a single proto-language at the root of the tree.

An article published in
Science
pursued this possibility and came up with a positive result. According to the article, there is a single source for human language, in Africa.
⁶¹
The evidence consisted of two main findings. First, it turned out that the place where phonemic diversity is greatest is Africa. The number of languages spoken on a
per capita
basis is higher in Africa than elsewhere. Second, phonemic diversity changes systematically as a function of physical distance from a region that,
a priori
, was unknown but,
a posteriori
, could be identified. That region turned out to be somewhere on the African continent. The exact location in African couldn’t be pinpointed, but that’s a detail not worth worrying about here.

Before this study was done, it wasn’t obvious that Africa would be where phonemic diversity would be greatest. Nor was it obvious that a site in Africa, or anywhere for that matter, could claim to be the seat of language. The analysis I’ve just mentioned suggested that languages have a single origin, as do species.
⁶²

Does the successful modeling of language evolution in Darwinian terms imply that the best possible way of modeling cognition within individuals is
Darwinian? Not necessarily. Some other way might prove better. Still, the success of Darwin’s approach in the domain of language, coupled with its success in culture and genetics, is uncanny. I doubt the same story could hold in so many domains by chance alone.

Because my aim in this book has been to advance a general theory of cognitive psychology, it’s worth checking whether the jungle principle is the right kind of theory. To address this question, it may help to recognize that theories can be said to occupy an abstract two-dimensional space (
Figure 16
). One dimension of the space, along its abscissa or
x
-axis, is its number of assumptions or free parameters. The other dimension, along the ordinate or
y
-axis, is how well the theory predicts data or, in the parlance of statistics, accounts for the variance (variability) in the data. By “data” I don’t just mean the data for which the theory was developed. Rather, I mean all the data that ultimately need explaining.
⁶³

It’s useful to divide theory space into quadrants. One of the four quadrants, on the lower right, is a
bad
quadrant. This quadrant is bad because it has slews of free parameters and assumptions.

Two other quadrants are
mediocre
. In one, the upper right corner, the number of free parameters and assumptions approximates the number of points to be predicted. This region is mediocre because no serious investigator wants to offer a theory with as many distinct theory elements as data points to be explained.

The other mediocre quadrant, the lower left corner, has few free parameters or assumptions, but little variability is accounted for. Being in the lower left corner of the space is like being where you were when you started. At this
early stage, you could make few assumptions because you had no idea what assumptions to make, and you could predict little or nothing.

FIGURE 16.
Theory space, with the holy grail shown in the top left corner.

Where you’d like to be is in the
good
quadrant, the upper left corner of the space. Here there are few assumptions or free parameters, and all or most of the variance is explained. The ideal point in the good quadrant is the top left corner. That point is the holy grail of science. It is where there are no assumptions or parameters but you explain everything.

Knowing where the good part of theory space is has led cognitive psychologists to limit the range of phenomena they’ve studied. They’ve limited their focus on the grounds that explaining all or most of the variance in a narrow range is at least suggestive of the possibility that the explanation is on the right track for explaining all the variance in the wider space.

Consider a hypothetical cognitive psychologist who spends his or her career studying how people judge whether one line is longer than the other. As it happens, I know of no such cognitive psychologist; the example is purely hypothetical.

By pursuing such a limited problem, the hypothetical investigator might be assured of developing a theory that occupies a relatively good part of a restricted theory space. He or she might investigate various facets of the problem: How does line-length discrimination depend on the colors of the lines, the thicknesses of the lines, where the lines are positioned in the visual field, how the lines are oriented, how long the lines can be inspected, whether the lines’ positions are predictable, and so on. The researcher might be able to predict his or her data with reasonably high accuracy with a theory that is admirably simple. But would this be satisfying? Should a cognitive psychologist limit what s/he studies to increase the odds of yielding an outcome that only hints at where s/he would like to be in the larger theory space that we all care about ultimately?

Judging from the thousands of studies published in cognitive psychology journals, it’s clear that many cognitive psychologists take this less-is-more approach.
⁶⁴
Going to research conferences confirms this impression. This is where cognitive psychologists usually present their findings before submitting their manuscripts to research journals. Cognitive psychologists, like other researchers, want to have nice-looking data to show their peers.

A defense of this strategy is that you have to start somewhere. You might want to “explain everything,” but that’s hopeless. Better to begin with some manageable problem and make some headway with it.

The problem is that cognitive psychologists have been “starting somewhere” for years. Thousands of studies in cognitive psychology occupy the
part of theory space where, while much of the variance within a narrow range is accounted for, very little variance is explained out of all the variance that ultimately needs explaining.

Studying things on a small scale is justified, of course, if the truths they expose are more broadly illuminating. Still, the number of cognitive psychologists who relate their focal questions to broader principles is small. Few investigators have the courage to offer grand, encompassing theories, or even to remark on the implications of their results for larger concerns. Their reticence is understandable given that espousing a grand theory or going beyond one’s own results can put one at risk of occupying (or seeming to occupy) a mediocre or bad part of theory space. Why venture from a safe place if you don’t have to?

Something interesting happens, however, if you look at the range of theories, models, and hypotheses that cognitive researchers report in research journals and at conferences. Their accounts tend to use similar terms and make similar assumptions. The particulars of the theoretical claims change a bit from study to study, but virtually all the claims rely on the same primitives: internal competition and cooperation among elements whose identities or strengths change over time. Virtually without exception, cognitive psychologists end up with an account that, in broad strokes, amounts to the claim that it’s a jungle in there.

Realizing this, I managed to overcome the feeling of slight disappointment I had as I attended talks at conferences or as I read papers in academic journals, including the one I had the privilege of editing from 2000 to 2005, the
Journal of Experimental Psychology: Human Perception and Performance
.
⁶⁵
Attending research conferences and reading published or as yet unpublished articles often gave me the feeling that, despite surface differences, most people were saying more or less the same thing—that inside the brain, there are components that cooperate or compete, and that out of these dynamics come the phenomena of cognition. I decided to write this book when I recognized this common view and came to believe that the commonness of the view is not so commonly appreciated.

Identifying commonalities is like boiling things down to basics. Have I boiled things down too much? If you boil vegetables too long, you can end up with mush. Have I taken cognitive psychology too far beyond the boiling point?

For some reading this book—students as well as professionals—the answer may be “Yes, the theory is too broad.” If you prefer exact quantitative accounts of narrow data sets, you may not take kindly to the generalization offered here. Hopefully, though, even if you’re in this camp, you may appreciate that
my aim has been integrative. I’ve tried to present the material so it comes across as a coherent whole, not a hodgepodge of vaguely related findings.

There have been attempts at unification before, of course, including rallying cries meant to corral the findings of cognition. One such attempt was the proclamation that we go beyond the information given.
⁶⁶
Another was the assertion that radical behaviorism was wrong. These are accepted statements now. If there is some other overarching theme for cognitive psychology, I don’t know what it is.

If the theoretical framework presented here doesn’t provide an exact quantitative account, it’s worth asking whether such an account is actually the one you should always pursue. One way to approach this question is to return to theory space and ask how you would judge how good a theory is.

A first answer is to check which quadrant of theory space the theory occupies. Obviously, if it’s in the bad quadrant it’s bad. Similarly, if it’s in a mediocre quadrant it’s mediocre, and if it’s in the good quadrant it’s good. Beyond this, you can ask how far a theory is from the optimal point in theory space. The optimal spot—the holy grail of theory development—is the upper left corner, as mentioned earlier. That is where there are no assumptions and no free parameters, but all the variance is explained. The holy grail can’t be obtained, of course, which is why it has the name it does. Still, like any holy grail, it’s useful to imagine it as a goal to which you can aspire.