Wednesday, August 31, 2005


New Orleans is one of my favorite places in the world, and I've met many incredibly nice and interesting people (with worse accents than mine ever was) there. It's been years since I've been to Mississippi or Alabama, but I have fond memories of those places too. It's painful to see the pictures of the devestation, and I can't help but feel the deepest sympathy for all the people affected, especially those who will have a difficult time dealing with and recovering from this tragedy financially. There is a great deal of poverty in southern Louisiana, Mississippi, and Alabama, and I'm sure that this will be too much for many to overcome on their own. I hope everyone who can is donating to the Red Cross, or another equally reputable relief organization, so that no one will have to get through this alone.

CogBlogGroup: Tomasello, Ch. 2, Part 2: The Human Ratchet

The third section of Chapter 2, beginning on p. 37, is where Tomasello really begins the book. Up to this point, he's been arguing that his hypothesis is viable, based on the differences in human and nonhuman primate social cognition. Now he begins to show what his hypothesis can give us. And he doesn't start small. He goes straight to language and mathematics, to of the most striking human achievements. And along the way, he takes a stand that, in my view, makes this an excellent book. But I will get to that in a moment. First, let's look at the ratchet effect and the sociogenesis of language and mathematics.

By page 37, Tomasello feels that he's given us enough evidence to convince the convinceable that one of the major cognitive differences between humans and nonhuman primates lies in our ability to understand others as intentional agents like ourselves. This understanding makes possible, among other things, imitative learning, and as he puts it, imitative learning gives us the ability to learn new behaviors by understanding the goals of the behavior, allowing us to "faithfully [preserve] newly innovated strategies" (p. 40, and this in turn puts us "into a new cognitive space" (p. 39). This placement into a new cognitive space is important, because creativity is dependent on existing knowledge. Innovations always need something to build on, and by learning more than mere associations through social interactions, we are provided with the building blocks for further innovations. A tool one person makes can be improved upon by the person who learns to use that tool from her, and then that tool can be improved upon, and so on, and so on, until our knowledge-base grows to the point at which we get the sorts of technological explosions that humans have seen over the last few millennia, and over the last century in particular. This process is further enhanced by the ability to work in direct collaboration with others, in essence pooling our knowledge resources, and producing even better innovations. The process of learning and then improving upon the innovations of those who come before us is the ratchet effect, and the two types of cooperation that I described, building, individually, on something that someone else has created, and building on something in direct collaboration with another individual, Tomasello calls virtual and simultaneous collaboration.

As an example of virtual collaboration, Tomasello uses the development of the collections of linguistic symbols and constructions that make up individual languages. While children initially learn fairly exact copies of the symbols and constructions that their parents use, languages develop over time, meaning that as new individuals learn them, they make alterations (most often unconsciously, I'm sure) that end up making those symbols and constructions more efficient for speech and communication. On pages 43 and 44, he gives several examples of how English expressions have become more efficient over time, usually through shortening the expressions (by removing redundancies, e.g.). This is just sort of collaborative construction process that the model of Hutchins and Hazelhurst simulates. Their model shows just how powerful such collaboration is in the creation of a shared system of linguistic symbols. For instance, their model shows how different systems of symbols can develop by dividing agents into groups that can only collaborate with other members of their groups.

While there are many languages, language itself is shared by all human cultures. This may be, as Tomasello notes, because some aspects (perhaps its underlying structure) of language are innate. This is the Chomskyan position, and Tomasello is quick to point out that his hypothesis is not inconsistent with some versions of that position (he mentions the principles and parameters approach, which I briefly described here). But it is also consistent with the creation of a simple symbol system by our earliest human ancestors some 250,000 years ago. Out of this system, all languages would develop through the sort of collaboration Tomasello as well as Hutchins and Hazelhurst describe. Similarities across languages in their structure, which are fairly common as linguists will tell you, could then be a product of the fact that languages develop to work with existing cognitive and physical (e.g., the anatomy of the vocal tract) mechanisms.

After language, Tomasello describes how the sociogenesis of mathematics might have worked. Mathematics, unlike language, is not universal, and while some have posited that we have innate arithmetical abilities (the addition and subtraction of small numbers, specifically), few have argued that we have a complex mathematical cognition module. Instead, mathematical knowledge develops using existing cognitive mechanisms, and only to the extent that it is needed in a particular culture. As cultures develop more complex economic, architectural, and technological systems, mathematical knowledge grows more complex. And often, particularly at the highest levels, mathematics develops through the simultaneous collaboration of small groups of individuals (people we would call mathematicians). Each innovation along the way is built upon previous innovations, and thus the development of mathematics is a particularly impressive example of the ratchet effect.

After the discussion of language and mathematics as examples of the power of sociogenesis, Tomasello moves on to human ontogeny (human development), and he begins this discussion, which will comprise the bulk of Chapter 3, with a philosophical discussion. And I must say that this short discussion (in the subsection titled "Philosophical Nativism and Development," beginning on p. 48) is one of the main reasons I like this book so much, and why I think it's a very good first book for the reading group. This is because he touches on one of the foremost debates in cognitive science: the nativism vs empiricism debate. Since its inception in the mid-1950s, cognitive science has been dominated by a few far-reaching debates, the most famous of which being the imagery debate, connectionist vs symbolic architecture, and nativism vs empiricism (there has also been a debate over the use of similarity-based vs rule-based processes within the symbolicist camp, but this is more local and technical, and therefore less famous). The imagery debate was an incredibly heated debate in the 1970s over whether representational primitives (the building blocks of all representations) were images or propositions. This debate died out in the early 80s, with most cognitive scientists deciding that both images and propositions both comprised our representational primitives, but with an emphasis on propositions. Recently, the debate has been making a comeback, largely due to the work of Larry Barsalou, who argues that most or all of our representations are images (which need not be visual, I should add). The connectionist vs symbolic architecture debate, which took off in the 1980s (perhaps because people were sick of the imagery debate) has been equally heated, and concerns the ways in which our brains process information. Do they do so using syntactic processes (rules) operating over discrete symbols (symbolic representations), or do they use analog representations that allow for similarity-based pattern recognition, as in connectionist models? Despite more than two decades of fighting, neither side has one out yet, and I wouldn't be surprised if this debate ended in a compromise like the one that initially ended the imagery debate.

But the nativism vs empiricism debate has been a part of cognitive science since its birth, and probably isn't going away anytime soon. As you probably know, cognitive science was born out of a revolt against behaviorism. Behaviorism was a particularly strong version of empiricism, in which all, or virtually all behavior was learned through associations. Many of the cognitive revolutionaries rebelled against empiricism as they rebelled against behaviorism. The most famous of these is, of course, Noam Chomsky, whose nativist theory of language has dominated linguistics and cognitive science in general for nearly 50 years. But empiricism didn't die with behaviorism. So for those same 50 years, people have debated whether pretty much every particular cognitive ability and process is innate or learned. For the most part, cognitive scientists now believe that some processes are innate, and some are learned, and no one is a strict nativist or empiricist (except Jerry Fodor, who is a strict nativist, believing that even concepts are innate). But there are still local debates over innateness, and in most cases, a particular cognitive ability is thought to be either innate or learned, with no middle ground.

This is where Tomasello takes issue with the way cognitive scientists view the mind. He argues that this either/or way of approaching things is bad biology, and it's not very useful either. Very few of our cognitive abilities are present, in their entirety, at birth (though if they are, or even if primitive versions are available early in infancy, nativists will argue that they are innate). Instead, these abilities develop over time and through our interactions with our environment. And this, Tomasello believes, is what we should be trying to understand: how these abilities develop ontogenetically. Trying to decide whether the abilities are innate or not doesn't help us in that task, unless it is directed at attempting to understand that development. I think this is a very important message, and one that I obviously haven't fully taken to heart (in a paragraph in this post, I've taken an either/or perspective, even). But it is a message that we cognitive scientists must learn if we're going to be able to use the study of our cognitive development to fully understand human cognition.

Now, I can't say that I find Tomasello's approach to the study of development entirely satisfying. His approach is to divide learning into two types: individual and cultural. Individual learning occurs through insight that we gain while interacting with our environment on our own, while cultural learning occurs through collaborative interactions with other individuals. This may be a fruitful approach experimentally, in that it may be important to control either individual or cultural factors, to the extent that it is possible, in order to understand particular factors in cognitive development. But in practice, and certainly in actual development, I'm not sure it's really possible to separate the two. Sure, human children are occasionally alone (some more than others), but for the most part, our development takes place in a rich social environment, and while not all of our interactions with that environment are collaborative, they are all impacted by culture. Where does cultural learning end, and individual learning begin? I'm not sure cognitive science can provide an answer to that question, or that it needs to.

Still, I think Tomasello's general approach, which is to focus on development, be it individual or cultural, is the right one. God knows how many hours of research and pages in journals have been wasted arguing over whether a particular cognitive ability is innate, and with little insight into the ability itself coming of it. If we focus on development, we can gain great insight into the different aspects of those individual abilities. The book chapter on causal reasoning in infants that I linked in the last post on Tomasello (it's here, if you didn't read it then) is a great example of this. By looking at the stages of the development of infants' concepts of causation, the authors show us different aspects that are contained in our fully developed causal reasoning. Of course, if you read the chapter through, you'll notice that the authors end up taking a stand on one side of the nativism-empiricism debate, which is unnecessary, but old habits die hard. Fortunately, that wasn't the main focus of the chapter, and as a result, their review of the developmental research is enlightening.

In the next chapter, Tomasello will focus exclusively on development. He'll occasionally mention nativist and empiricist theories, and I must admit that at times it looks like he's taking a distinctly empiricist position, even if he believes that doing so is unproductive. But for the most part, his focus is on the developmental sequences in cognitive development. I hope his approach will serve as a model for all of those cognitive scientists who are stuck in mostly fruitless debates over innateness.

Monday, August 29, 2005

Two Good Resources on Language Evolution

One of the central topics in Tomasello's The Cultural Origins of Human Cognition is language -- how it evolved, how it develops ontogenetically, and how it enhances our cognitive abilities. But Tomasello discusses language at a fairly high level (e.g., syntax, semantics, and pragmatics). For spoken language to have evolved (biologically or culturally), several changes in the morphology of our vocal tracts had to take place. In addition, speech must be designed to take advantage of certain properties of our auditory systems, some of which may have evolved since the evolutionary lines of humans and other primates diverged. So, in order to really understand the language faculty, and how it evolved, you have to understand speech production and speech perception. With that in mind, here are two good papers on the internet.

The first is a book chapter titled "What Are the Uniquely Human Components of the Language Faculty?" by Marc Hauser and Tecumseh Fitch. Yes, that Hauser and Fitch. But this chapter isn't about the recursion hypothesis. It's a nice historical review of the speech production and perception literature. Unfortunately, since it's a book chapter, the scanned version doesn't include the references, but if they cite something that you think you'd like to read, and you can't find the full reference, let me know, and I will find it for you.

The second source, which is cited by Hauser and Fitch, is a Behavioral and Brain Sciences paper by Peter MacNeilage titled "The Frame/Content Theory of Evolution of Speech Production." It's mostly about speech production, as the title suggests, and contains a lot of great info on the physical aspects of speech, along with a theory about how these and neural organization influenced the evolution of language.

Dennett to ID: Get in Line

If you haven't read this opinion piece by Daniel Dennet in the New York Times (via The Fly Bottle), then I highly suggest you do so now. It may be the best opinion piece on Intelligent Design that I've seen yet, not because he says anything new, but because he distills the issue down to its essence and leaves it at that. When he does so, Intelligent Design is shown in its true light, which is to say, a light shining on a scientifically empty idea. After I read it, I tried really to choose the passage I wanted to quote here, but each paragraph is really good, so it's damn near impossible to decide. So I'll just pick a passage semi-randomly:

Instead [of doing real science], the proponents of intelligent design use a ploy that works something like this. First you misuse or misdescribe some scientist's work. Then you get an angry rebuttal. Then, instead of dealing forthrightly with the charges leveled, you cite the rebuttal as evidence that there is a "controversy" to teach.

Note that the trick is content-free. You can use it on any topic. "Smith's work in geology supports my argument that the earth is flat," you say, misrepresenting Smith's work. When Smith responds with a denunciation of your misuse of her work, you respond, saying something like: "See what a controversy we have here? Professor Smith and I are locked in a titanic scientific debate. We should teach the controversy in the classrooms." And here is the delicious part: you can often exploit the very technicality of the issues to your own advantage, counting on most of us to miss the point in all the difficult details.

William Dembski, one of the most vocal supporters of intelligent design, notes that he provoked Thomas Schneider, a biologist, into a response that Dr. Dembski characterizes as "some hair-splitting that could only look ridiculous to outsider observers." What looks to scientists - and is - a knockout objection by Dr. Schneider is portrayed to most everyone else as ridiculous hair-splitting.

Now go read the rest.

CogBlogGroup: Tomasello Chapter 2, Part 1: What the Apes Got

Since Chapter 2 is pretty dense, I'm going to talk about it in two separate posts. The first is about chimpanzees (mostly), and the second about humans.

Tomasello begins Chapter 2 with the distinction that drives the entire book: the distinction between biological and cultural inheritance. In fact, he states the hypothesis of the book in two sentences just a few paragraphs in to Ch. 2:
My particular claim is that in the cognitive realm the biological inheritance of humans is very much like that of other primates. There is just one major difference, and that is the fact that human beings "identify" with conspecifics more deeply than do other primates. (p. 14)
To argue for this thesis, he has to answer two questions, which he gives on p. 15:
  • How does the cognition of primates differ from that of other mammals?
  • How does the cognition of humans differ from other primates?
The first of these is important, because if Tomasello's hypothesis is correct, then understanding the differences between our closest evolutionary relatives, other primates, and other mammals will help us understand which aspects of human cognition developed through biological inheritance. Answering the second question will then serve two purposes. First, if he can demonstrate that humans can "'identify' with conspecifics more deeply than do other primates," then he can argue that other cognitive differences between humans and other primates may have developed through cultural transmission. Second, understanding the differences between human cognition and the cognition of other primates, over and above differences in theory of mind, will allow Tomasello to determine what it is, exactly, that may have developed through cultural transmission. The primary purpose of Chapter 2 is to do the first of these. He spends much of the rest of the book arguing that the demonstrated differences in theory of mind can in fact lead to the other cognitive differences we observe between humans and primates.

So what can most mammals do, cognitively? Tomasello gives two lists, the first general and the second social (p. 16-17). I'll just discuss one, because he doesn't explain it very well. On the first list (non-social cognitive abilities), he includes the ability to pass object permanence tests. What is an object permanence test? It's just what it sounds like. It's a test to determine if an individual (human or nonhuman) understands the basic properties of solid objects, which include numerosity (most things don't spontaneously multiply themselves), motion, and permanence (things don't just disappear). Here are some examples of classic Piegetian object permanence tests (from this paper):

These were obviously designed for children. In a typical experiment, the child, often a very young infant, will be shown several different versions, some of which an individual would expect if she understood the properties of solid objects, and some of which she wouldn't. Experimenters determine whether the child understands these properties by measuring how surprised the child is when a particular event occurs (often by measuring looking time or sucking, both of which will increase if the child sees something novel or unexpected). For example, in the top two sets of drawings above, an object (in this case, a toy train) starts out in plain view, and then disappears behind a larger solid object (the rock). If the train does not come out the other side, then the child should expect the object to still be behind the rock when it is lifted. If the train has come into view on the other side of the rock, the child should be surprised to see a train when the rock is lifted. Thus, if a child looks longer in the second case than in the first, we can infer that the child understands that particular aspect of object permanence. Human infants understand this from a very, very young age, and other mammals appear to understand it as well.

Back to Tomasello. After listing many similarities between primates and other mammals, he singles out one particular difference: the ability of primates to understand relational categories, particularly in the form of "third-party social relationships" (p. 17). While most mammals are able to understand kinship and dominance relationships between themselves and other individuals, primates are able to understand kinship and dominance relationships between two other individuals, and to use this knowledge in interacting with those individuals. Primates, unlike other mammals, are also able to understand relational categories in non-social domains, though as Tomasello notes, they must be extensively trained to do so. In essence, you could say that primates, but not other mammals, developed a concept of "same" in the social realm, and with extensive training, they can extend this concept to non-social domains. For example, if you give a dog two objects, one of which is larger than the other, and reward it for picking the larger object, it will eventually pick the larger object all of the time. If you then substitute the first two objects for two different objects of different sizes than the first, one of which is larger than the other, the dog will not know to pick the larger object until it has been trained to pick that specific object. It doesn't understand that the two pairs of objects involve the same relationship: one is bigger than the other. Chimps, on the other hand, can be trained to do this, though it takes a while. Human children, on the other hand, do this early in development and without any training.

This is where Tomasello draws the line between primate cognition and the cognition of other mammals. He then begins to draw the line between human and primate cognition. The first difference he notes is human understanding of causation. And it is this understanding that allows humans to attribute beliefs and goals to other humans. Only after gaining the ability to represent causal relations between two events (rather than simply associating the two events) can we understand that an individual's beliefs or goals caused him or her to act in a certain way. Interestingly, just as Tomasello believes that primate understanding of relational categories in non-social domains is an extension of their understanding of such categories in social domains, he believes that human understanding of causal relations in general arose from our understanding of mental causation.

I have to admit, I'm doubtful about that. Whereas chimpanzees have a very difficult time extending their social understanding of relational categories to physical relational categories, human infants appear to develop a concept of causation independently of the development of theory of mind. While there isn't much evidence for a "causation" module in the human brain, infants do begin to develop an increasingly complex concept of causation beginning at about six months of age, which is long before most of our theory of mind capabilities begin to emerge (check out this book chapter for a great review of the literature on infant causal knowledge). If our causal knowledge were dependent on our knowledge of mental causation, we would expect the latter to develop before, or at least simultaneously with causal knowledge.

But I'm digressing again. From page 26 on, Tomasello discusses primate culture. By this time, you've probably realized that "culture" is a pretty loaded word. Many anthropologists and biologists use it to refer to something like the following1:
[B]ehavioral patterns that diffuse across a group of individuals through some form of social learning and is subsequently displayed by a significant membership of the group. (p. 55)
Tomasello, on the other hand, reserves the term for cumulative culture that is achieved through imitative learning and direct teaching. And this is where we further see the dividing line between humans and nonhuman primates.

If we use the first definition of culture, then many nonhuman species can be said to display cultural cognition. Chimps, bonobos, macaques, and even dolphins and whales (as Clark notes) can develop behavioral patterns, ranging from tool use to courtship rituals, that are specific to particular isolated groups, and which are learned by young by observing adults (see this paper for a nice review of different types of cultural behaviors, in the broader sense, in chimps). But Tomasello argues that these aren't instances of what he refers to as culture, because they are not transmitted through teaching and imitation. Instead, they are transmitted through emulation learning and ontogenetic ritualization. Both of these, he notes, require a great deal of intelligence, but neither constitutes true cultural transmission.

Here is the example that Tomasello gives of emulation learning:
If a mother rolls a log and eats the insects underneath, her child will very likely follow suit. This is simply because the child learned from the mother's act that there are insects under the log -- a fact she did not know and very likely would not have discovered on her own. But she did not learn from her mother how to roll a log or to eat insects: these are things she already knew how to do or could learn how to do on her own. (p. 29)
The distinction between emulation learning and true imitation is very important. As Tomasello notes, there have been few, if any, observed cases of nonhuman primate parents (specifically mothers) teaching behaviors to their children. Instead, the mother performs a behavior with her child nearby, the child observes the end result (in the example above, the exposing of food), and repeats the act not because it's what the mother did, but because it understands the affordances of the objects involved, and uses that understanding to achieve the same end result that its mother did. This requires a very complex understanding of affordances, which, in the words of Norman Donald, are:
[T]he perceived and actual properties of the thing, primarily those fundamental properties that determine just how the thing could possibly be used. A chair affords ("is for") support, and, therefore, affords sitting. (p. 9)
It is through the complex understanding of the affordances of objects that nonhuman primates are able to innovatively develop tools. In order to use a stick to retrieve termites from a mound by placing it into holes in the mound, a chimpanzee must understand that a stick of a certain size will fit into holes of a certain size, for example. But because chimpanzees are not learning to use tools from their conspecifics through imitation, but rather through emulation learning, their ability to develop and use tools is highly limited. This is reflected in the differences between human and nonhuman primate tool use, which Marc Hauser and Laurie Santos list in the following table2:

Notice that both nonhuman primates and humans are able to recognize the "relevant design features" and use "multiple materials for different tools." Both of these require understanding the affordances of objects, but little else. However, the rest of the abilities on the list, which only humans exhibit, require more than this. They require an understanding of causation (for developing tools that consist of "parts with different functions," for example), the ability to build upon existing uses of a tool (cumulative culture), and most likely, the ability to learn the use of a tool through something more than the mere association of an event and an outcome that is involved in emulation learning. If you only learn an association between an event (e.g., rolling a log or putting a stick in a termite mound) and an outcome (getting food, e.g.), it will be much more difficult for you to use the object or action to achieve other, perhaps unrelated outcomes.

The second type of social learning that Tomasello attributes to chimps is ontogenetic ritualization (p. 31-33). He describes this type of learning in the context of gestural communication. Nonhuman primates develop group-specific, or even individual-specific gestural signals through a process in which one or more individuals comes to associate a particular gesture with a particular outcome. Once again, this type of learning does not require imitation or an understanding of the goals of the gesturer/observer. Thus he writes
[Emulation] learning and ontogenetic ritualization are precisely the kinds of social learning one would expect of organisms that are very intelligent and quick to learn, but that do not understand others as intentional agents with whom they can align themselves. (p. 33)
Tomasello argues that emulation learning and ontogenetic ritualization can explain all of the group-specific behaviors observed in several primate species. While he doesn't address the issue, because he is more concerned with our closest evolutionary ancestors, these types of learning might also explain what many see as cultural learning in other mammal species, such as whales and dolphins. For an example, in an article linked by Clark in a post on Chapter 1, researchers report observing orcas develop a unique technique for capturing gulls. Once one whale started using this technique, the rest of the whales in the tank picked it up and began using it successfully as well. This sort of learning is consistent with other examples of cultural transmission (in the broader sense) observed in toothed whales (see this paper for more examples). But at this point, there's no reason to think that the orcas were learning through imitation rather than emulation in which they observed a behavior and an outcome they desired, and thus emulated the behavior to achieve the same outcome. It may turn out that orcas, unlike nonhuman primates, are able to learn through imitation, which might present an interesting case of convergent evolution (and wouldn't affect Tomasello's arguments directly), but at this point there doesn't appear to be any reason to believe that's the case.

At the end of the section on primate culture in Chapter 2, Tomasello feels that the evidence overwhelmingly indicates that chimpanzees are not able to transmit information culturally (in his sense of the word), and that this is because they are not able to understand others as intentional agents. Thus, he feels he's answered both of the questions I quoted at the beginning of this post, and that the answers support his hypothesis. In the next post on Chapter 2 (which will be much shorter, I promise), I'll talk about his discussion of human culture. Hopefully by the end of Chapter 2, everyone will have a good understanding of just where Tomasello is going, and why.

1 Sinha, A. (2005). Not in their genes: Phenotypic flexibility, behavioural traditions cultural evolution in wild bonnet macaques.Journal of Biosciences, 30(1), 51-64.
2 Table 1 (p.20) in Hauser, M.D., & Santos, L.R. (In Press). The evolutionary ancestry of our knowledge of tools: From percepts to concepts.

Sunday, August 28, 2005

Argh Again

This post is a personal rant. Please ignore it.

It's been a busy couple of weeks, what with the beginning of the semester and all, and for the last few days, my son has been sick (sick and grumpy), so I've been way behind on my blog reading for some time. Today, my son is with his mother, and I'm just sitting around, so I decided to catch up on my blog reading. In the process, I came across this post by Brian Leiter from two weeks ago. It's an 1100+ word rant on a review of Friedrich Nietzsche by Curtis Cate in the New York Times. If you're a reader of Leiter, you probably know that 1100+ words is pretty long for a post that Leiter has actually written (rather than mostly quoted). And he is a Nietzsche expert, so I was eager to read it.

The gist of the post is that Leiter doesn't think that the author of the review, William T. Vollman, knows what he's talking about, and he provides several examples of Vollman's ignorance to prove it. As I read it, I was particularly struck by this in the last paragraph:
The broader issue, though, is about the responsibilities of newspapers, like The New York Times, that aspire to be serious and intellectual. If a decision is made to commission a long review of a book, why not enlist someone who actually knows something?
Now that's a sentiment with which I can agree wholeheartedly, as anyone who's read this blog knows (and hey, I told you this was a personal rant, so why are you still reading it?). But as I read it, something didn't sit right with me. It's not that I think Leiter's wrong about Vollman's knowledge of Nietzsche, it's that I don't think Leiter really believes what he says about the "responsibilities of newspapers." I mean, I think the thinks he believes it, but I don't think he really believes it. You see, when I read it, my mind immediately went back to this post . In it, Leiter favorably links to this review of David Buller's Adapting Minds, by Sharon Begley. So favorably, in fact, that Leiter wrote:
Law-and-economics folks, who are often especially partial to this shoddy science, would do well to read the review and the book. [Emphasis mine]
But wait a second. Wait a friggin' second! That review was written by someone with no qualifications in psychology, or philosophy, and it shows. She's a science editor for the Wall Street Journal, and has written on all sorts of scientific topics (ranging from psychology to astronomy), but she's not an expert in any of the areas of psychology or biology that Buller covers. To give one example of the ignorance of the issues that Begley displays, the last sentence of the review read:
After "Adapting Minds," it is impossible to ever again think that human behavior is the Stone Age artifact that evolutionary psychology claims.
Ummm... no. Buller's book is very good in places, especially where he sticks to existing critiques of Evolutionary Psychology, but it fails where a critique of EP must be at its strongest, in critiquing the work of Cosmides and Tooby (see the link under "no" for why this is the area in which it is most important to critique EP, and why Buller fails to do so effectively). And thus, contrary to the title of Leiter's post and the last sentence of Begley's review, Buller has not demolished Evolutionary Psychology. Anyone who was qualified to write that review would have known that, as would anyone who was qualified to evaluate the book and the review.

When I pointed this out, and criticized Leiter for endorsing the review and hyperbolically claiming that the book itself spelled the end of Evolutionary Psychology, Leiter was none too happy, and he made this very clear in an email exchange. All of this begs the question, why does Leiter have no problem endorsing a book review by someone who is not qualified to write it, to the point of being rather angry about any criticism of that endorsement, in one instance, but he writes an extended post ending with a criticism of newspapers that use unqualified reviewers in another? My best guess is that Leiter endorsed Begley's review because it came to the same conclusion he had, and that he cared not about the substance of the review itself, but when the NYT's published a review on a subject that is near and dear to him (he's written a book on Nietzsche, as I'm sure you know), the substance of the review suddenly became important to him. My guess may be wrong, but I can't think of any answer to that question in which Leiter comes out looking good.

OK, thus ends my rant. I'll admit that it's probably unhealthy for me to have spent that much time on it, but this sort of thing really bugs me. I mean, Leiter and I agree about the merits of Evolutionary Psychology, even if the argument he has presented against it isn't a very good one. Heck, Leiter and I probably agree on a lot of things, as we're both on the same end of the political spectrum. Though I must say it's hard to tell, because it's rare that Leiter actually offers any positive political positions on his blog, and his scholarly writings aren't overtly political (at least not the ones I know of). But regardless of whether we ultimately agree or disagree, Leiter's selective criticism of intellectual irresponsibility strikes me as just another instance of the attitude that leads to the politicization of science and other scholarly endeavors. You just can't decide which scholarship is good and which is bad based on whether you agree with the conclusions. Otherwise, you're no better than the creationists and global-warming "skeptics," even if the issues are less pressing.

Saturday, August 27, 2005

When Science, and the Philosophy of it, Meet Blogs and the Press

I feel sorry for Elizabeth Lloyd. I haven't read her book, as I'm not really up on all of he "evolution of the orgasm" literature, so reading a critique of a particular perspective would be pretty pointless. But what is painfully clear is that she has been a victim of the inability of the press to understand science, and the further inability of people to look past the bad reporting and their own prejudices when evaluating science. If ever there was an argument for better science education and reporting, it is Lloyd's post over at Philosophy of Biology. She provides several examples of outrageous comments on her book and the ideas within it. I'll just give a couple:
  • And we all think Dr. Elisabeth Lloyd is something of an Uncle Tom for women.
  • Thinking women, liberal or conservative, will not be thrilled by this book. Fundamentalist Christians, on the other hand, will be ecstatic.
As these and the rest of Lloyd's examples show, conservatives aren't alone in their politicized ignorance of science. Lloyd notes that she didn't "cherry-picked the worst of the bloggers; these are nearly all intelligent and relatively scientifically-informed feminists who are keenly worried about what theyhave heard." I think Lloyd's description of them as "scientifically-informed" may be a bit generous, but if it is accurate, then the quotes she attributes to them are that much worse, because they are made by people who should know better.

I've always believed that there are two general kinds of creationists: the ignorant and the stupid. The ignorant are your everyday, gardenvarietyy creationists, who would ask you things like, "If evolution is true, then why are their still moneys?" or claim that "Evolution violates the Second Law of Thermodynamics." These are people who know nothing about evolution (or thermodynamics, for that matter), and show it in everything they say about it. The vast majority of these people don't really care about evolution, and could go days or weeks without ever thinking or talking about it. For them, it's largely about religion and culture, and science is at most a scapegoat. Then there are the stupid creationists. These are the people who do have a little knowledge of evolution, or mathematics, physics, or chemistry, and who use it to actively campaign against evolution for largely political (notice I didn't say religious) and self-aggrandizing purposes. They're the ones who spend as much of their time as they can talking about evolution and science; who make sure that the ignorant remain that way, and that they think about evolution and science as much as they can, but from the stupid creationists' perspective. They really have enough knowledge to see where they are wrong, or at least to know that they don't know enough to make theclaim that they know what they're talking about. Feminists who are "scientifically informed," but still spout nonsense like the above, clearly without having read Lloyd's book, are no better than the stupid creationists, and just as dangerous as them. And just as in the case of creationism, the press is helping them along.

Tuesday, August 23, 2005

Traveling from West to East, Cognitively: The Causes of Cultural Differences in Reasoning

Since the publication of Richard Nisbett'’s book The Geography of Thought, researchers have been scrambling to study differences in cognitive style between Westerners (Europeans and Americans) and East Asians. In his book, Nisbett argues that Westerners reason more analytically, while East Asians reason more holistically, intuitively, and dialectically (meaning they're more likely to consider alternative views and take the middle ground). For example, East Asians prefer dialectical proverbs, in which there is a logical contraciction (e.g., "“Too humble is half proud") to proverbs in which there is no such contradiction (e.g., "“For example is no proof"”), while Americans prefer the latter. East Asians also prefer dialectical solutions to social problems that involve contradictions, like the following:
Mary, Phoebe, and Julie all have daughters. Each mother has held a set of values which has guided her efforts to raise her daughter. Now the daughters have grown up, and each of them is rejecting many of her mother's values. How did it happen and what should they do?
Americans, on the other hand, prefer solutions that donÂ’t contain a contradiction. Thus the East Asian participants in the experiment that used stories like the one above tended to lay some of the blame on both the mother and the daughters, while Americans tended to blame one or the other1. Nisbett and his colleagues have taken these findings to indicate that East Asians prefer dialectical to formal reasoning.

This preference for dialectical reasoning also manifests itself as a preference for intuitive over rule-based reasoning in other domains, such as categorization. Norenzayan et al. demonstrated this preference in a categorization task2. As an example, they use the following question: "“Is the Pope a bachelor?" If we adopt a rule-based approach, then we must conclude that the Pope is a bachelor because he is an unmarried adult male. However, if we take a more intuitive approach, we will tend to think about past examples of bachelors, and conclude that since the Pope isnÂ’t really very similar to those past examples, he is not a bachelor. When given questions like these, Americans tended to adopt a rule-based approach, while Chinese and Korean participants tended to categorize using a more intuitive, exemplar-based approach.

In addition to dialectical reasoning, East Asians Nisbett has argued that East Asians prefer holistic reasoning to analytical reasoning. This is most evident in situations in which East Asians rely more on context than Westerners, who prefer analytical reasoning. For example, Masuda and Nisbett3 conducted a memory experiment in which they presented Japanese and American participants with animated scenes. After a delay, they presented the scenes again, and asked the participants to answer whether an object in the scene was one they had seen before. Some of the objects were from the original scenes, and some were new. In addition, some of the old objects were presented in new scenes (i.e. a novel context). They found that Japanese participants made more errors when the object was in a novel context, while the Americans were unaffected by the context.

A just published study by Chua et al. demonstrates the increased attention to context at the level of eye-movements during the processing of scenes. John Hawks and Razib have already posted good descriptions of the study, so that I don't have to. I'll just say that the researchers found that Chinese participants tended to look at the background (context) much more than American participants, who tended to look at the focal objects more than the Chinese participants.

All of the work by Nisbett and colleagues, including the new eye movement study by Chua et al., do little more than demonstrate effects. There hasn't been much of an effort to test hypotheses about why these cultural differences in reasoning styles might exist. Nisbett has explained these differences as being the result of the dominance of individualism in Western cultures, and collectivism in Eastern cultures, but there is no experimental evidence to support this view. In fact, the very distinction between collectivist and individualist cultures may not be a very good one, as evidenced by research attempting to measure cross-cultural differences on these dimensions has shown4.

So what might cause these differences? Researchers have recently suggested that pervasive differences in social environments across cultures may play a role. In particular, they have suggested that fear of isolation, which manifests as a fear of being excluded from a group due to, among other things, failure to conform to majority opinions or trends, might affect reasoning styles, and they have shown that East Asians have higher levels of fear of isolation, on average, than Westerners5. But as they say, correlation is not causation. In order to really test the idea that fear of isolation causes differences in reasoning, you'd have to randomly assign participants to high or low fear of isolation conditions. That means experimentally manipulating fear of isolation. Then you'd have to show that inducing high levels of fear of isolation results in reasoning differences similar to those between East Asians and Westerners. So that's what they did6. They first randomly assigned Western (American) participants into one of two groups, and then had them write about past experiences. In the low fear of isolation group, they wrote about an experience in which they had isolated another individual from their group. In the high fear of isolation condition, they wrote about an experience of being isolated from a group. This caused the participants in the high fear of isolation group to score significantly higher on a fear of isolation test than those in the low fear of isolation group.

After manipulating fear of isolation levels, the researchers gave the participants tasks similar to those in the studies described above. In one experiment, they had them choose between dialectical and logically consistent proverbs, and tested the effects of context on memory. They found that participants in the high fear of isolation group preferred dialectical proverbs relative to the low fear of isolation group (they also showed that this preference correlated positively with individual differences in fear of isolation in an East Asian sample). In a second experiment, they gave participants a memory task identical to the one in Masuda and Nisbett, and showed that participants in the high fear of isolation group relied on context in a recognition memory task than participants in the low fear of isolation task. Thus, simply by manipulating fear of isolation levels, you can produce reasoning preferences in Westerners that resemble those of East Asians.

From these experiments we can conclude that fear of isolation plays a causal role in cultural differences in reasoning. Exactly how it affects reasoning was not studied, but it's likely that fear of isolation makes people pay more attention to multiple individuals (creating a preference for dialectical reasoning) and to the overall context. The authors of the studies note, however, that fear of isolation is likely not the whole story, and I suspect that in the future, researchers will look at other social factors in order to begin to piece together the causal picture. At the very least, though, these experiments rule out one explanation. Because we can induce Westerners to reason like East Asians, it's highly unlikely that there are large differences in the cognitive architectures of the two groups.

1 Both of these findings are from Sanchez-Burks, J., Nisbett, R. E., & Ybarra, O. (2000). Cultural styles, relationship schemas, and prejudice against outgroups. Journal of Personality and Social Psychology, 79, 174-189.
2 Norenzayan, A., Smith, E.E., Kim, B. J., & Nisbett, R. E. (In Press). Cultural preferences for formal versus intuitive reasoning. (In press). Cognitive Science.
3 Masuda, T., & Nisbett, R. E. (2001). Attending holistically versus analytically: Comparing the context sensitivity of Japanese and Americans. Journal of Personality and Social Psychology, 81, 992-934.
4 Kim, K., & Markman, A.B. (In Press). Differences in Fear of Isolation as an explanation of Cultural Differences: Evidence from memory and reasoning. Journal of Experimental Social Psychology.
5 Kim and Markman, In Press.
6 Kim and Markman, In Press.

The Return of Hauser, Chomsky, and Fitch

Only this time, it's Fitch, Hauser, and Chomsky. They're playing musical authors.

Anyway, recall that Hauser, Chomsky, and Fitch published a paper a few years ago in Science in which they argued that an explanation of the evolution of language may require only one new ability: recursion. In fact, recursion may be the only new ability we need to explain a whole host of cognitive skills that are unique to humans. You may think Tomasello oversimplifies the case in arguing that all we needed were some adaptations that allowed for collaborative learning, but HCF have brought oversimplification to a level not before witnessed in the behavioral sciences. Steven Pinker and Ray Jackendoff responded to this silliness with a very good paper of their own, in which they show that the HCF theory just doesn't hold water. If you're keeping up with this debate, you might be interested to know that HCF, now FHC, have written a response, which will be published in the same journal that published the Pinker and Jackendoff paper (thanks to Razib, who I should note is a member of the reading group, which automatically makes him cool, for pointing this out).

If you weren't convinced by the HCF paper, you won't be convinced by the FHC. It's more of the same, with a lot of time spent detailing what they perceive as misconceptions on the part of Pinker and Jackendoff. They do make one very good point, which I guarantee was insisted upon by Hauser (because he's written about it elsewhere), namely that speculation about the adaptive function of language at different points in its evolution is worthless. It's better to use comparative methods (comparing human linguistic abilities to the cognitive and communication abilities of nonhumans, especially monkeys and apes), and data from contemporary humans, because, well, that's real data, while idle speculation isn't. In addition, I do like their "faculty of language in the narrow sense" (FLN) and "faculty of language in the broad sense" (FLB) distinction. The FLN is just those aspects of human language that are unique to humans, and in humans, unique to language. The FLB is language in general, which can include properties that human languages share with nonhuman animals' communication systems, as well as aspects of language that are used by other cognitive systems, and thus not unique to language. I'm not so sure that Pinker and Jackendoff have mischaracterized this distinction, despite FHC's protests. But all in all, the paper is a good read, and I recommend it if you're interested in this sort of thing.

But beware! In the paper, there is a reference to an online appendix that discusses Pinker and Jackendoff's misconceptions about the Minimalist Program. Do not, if you value your sanity, read this appendix! It was obviously written by Chomsky, as it might as well have been written by the Chomskybot. It's no wonder, then, that the editor asked them to remove it from the published version of the paper. The damn things well nigh unreadable. Honestly, it amazes me that in an appendix designed to clear up misconceptions that are apparently based on a lack of requisite background knowledge on Pinker and Jackendoff's part, Chomsky writes an explanation that has holes large enough to drive an entire fleet of Mack trucks filled with background knowledge through. And on top of being about as clear as mud, the appendix is filled with Chomsky's typical ego-filled "where I go, so goes linguistics" rhetoric. When Chomsky ends the appendix with the statement, "Little [of Pinker and Jackendoff's critique] survives such analysis, as far as we can tell" (emphasis mine), I wonder if the "we" refers to Chomsky, with Hauser and Fitch just nodding their heads and saying, "Whatever you say, Noam. Whatever you say." Ugh.

I would love to hear from a linguist whether the appendix makes any sense to them. My suspicion is that for most, it won't. The rest of you, stay as far away from that appendix as is humanly possible. To borrow a phrase from Nietzsche, "When you gaze long into the Chomsky, the Chomsky also gazes into you."

UPDATE: Mark Liberman responds to my plea to linguists, writing:
I'll limit myself to observing that it's entirely "inside baseball": seven pages of text that mention no linguistic facts and no specific languages, nor any simulations, formulae, or empirical generalizations. Aside from a very general and abstract account of Chomsky's view of the goals of his research, the only topic is who said what when, sometimes with a very abstract explanation of why. It's an odd document -- I can't think of anything at all comparable from a major figure in a scientific or scholarly field, except perhaps some controversies over precedence (which is not an issue here). I agree with the judgment of Jacques Mehler, the editor of Cognition, who asked for it to be cut; and it seems to me that it's a distraction for outsiders (including most of the normal readership of Cognition) to try to understand it.
Now I don't feel so bad about making fun of it, even if the Nietzsche joke might have been a bit much.

Perhaps more importantly, though, Liberman links to the next turn in the debate over the original HCF paper, Pinker and Jackendoff (now Jackendoff and Pinker, not wanting to appear less egalitarian than Fitch, Hauser, and Chomsky) have written a response to the respone. It's here.

Monday, August 22, 2005

Autism and Theory of Mind

Over on Yahoo, we're discussing the role of autism in Tomasello's hypothesis in The Cultural Origins of Human Cognition. Tomasello, you will recall, argues that autism is an example of a disorder that results in a deficit in just the sorts of skills, which fall under the general heading of theory of mind, that he believes are responsible for most of the advances in human cognition because they make possible collaborative learning, and thereby culture. There are two ways in which using the example of autism might come back to bite Tomasello:
  1. If it turns out that some other deficit underlies the theory of mind deficits in autism, then Tomasello would have to revise his hypothesis to admit at least one other cognitive adaptation in humans.
  2. As Clark notes, many autistics are able to function quite well in non-social tasks. In fact, at the top end of the autism spectrum, individuals with Asperger Syndrome show the marked social deficits that characterize autism spectrum disorders, but are able to learn and reason in a way that is uniquely human, including developing language at a normal rate. There are people with Asperger Syndrome in physics departments and writing brilliant music, things that no nonhuman animals can do. If Tomasello is going to explain uniquely human cognitive abilities by reference to theory of mind and the cultural learning that it allows, then Asperger Syndrome and other high-functioning autistic individuals represent a wrench in his argument.
But discussion of those criticisms can wait until we're further along in the book. In this post, I'm not going to elaborate on them. Instead, I'm going to talk about some of the research on autism in cognitive psychology and cognitive neuroscience. In particular, I'm going to talk about the theory of mind theory of autism, and one of its alternatives, the weak central coherence theory. The research I'll describe certainly has implications for Tomasello's hypothesis, but that discussion will have to wait.

Most of the work on the relationship between theory of mind and autism has been conducted by Simon Baron-Cohen and his colleagues. Baron-Cohen (B-C) has theorized that there are two general cognitive styles: empathizing and systemizing. Empathizing, according to B-C, consists of a "drive to identify emotions and thoughts in others and to respond to these appropriately" (p. 302). Systemizing, on the other hand, involves "drive to analyze and build systems, with the aim of understanding and predicting non-agentive events" (p. 302)1. He believes that there are three general types of individuals: those who have strong empathizing skills and relatively weak systemizing skills; those who have strong systemizing skills and weak empathizing skills; and those who have balanced empathizing and systemizing skills. On average, males tend to be strong systemizers, while women tend to be strong empathizers. B-C has used this classification system to explain all sorts of behavioral differences between males and females, including differences in mathematical ability. While his theory that males and females exhibit differences in these abilities, on average, is highly controversial and, to say the least, not widely accepted among cognitive psychologists (see, e.g., this paper for a short critique of his empirical evidence), B-C has extended it to explain many of the symptoms of autism spectrum disorders. He believes that individuals with autism are extreme systemizers, and thus calls his theory the Extreme Mail Brain theory of autism2.

B-C has conducted what I consider to be thoroughly uninteresting experiments to test his systemizing-empathizing view of autism, using mostly Asperger patients. Here's an example experiment3: Asperger patients and non-autistic individuals, who have been matched for IQ (in other words, there's no IQ difference between the two groups) are given a systemizing and an empathizing task. The systemizing task involves making predictions in a simple folk mechanics task, while the empathizing task involves recognizing social faux pas in a set of social stories.

Pause here for a moment here, and look briefly at the diagnostic criteria for autism and Asperger specifically. You'll notice that one of the diagnostic criteria for Asperger (A. 3) is almost identical to B-C's definition of empathizing! Furthermore, one of the most common symptoms of both Asperger Syndrome and autism in general is an inability to perceive social nuances. I'll give you a real life examples. Asperger patients tend to have little knowledge of what is and is not appropriate in a given social context, and thus are unable to recognize faux pas. A doctor I know who works with Asperger patients once met with one for the first time, and in the course of their conversation, the patient used the word "fuck" constantly. He didn't know this doctor from Adam, and any non-autistic person would have avoided cussing in that context, but the Asperger patient used it over and over and over again. Another consequence of failing to pick up social nuances is an almost complete inability to recognize lies and figurative speech, and thus take everything literally. Here's another real life example. Another doctor I know was meeting an Asperger patient for the first time. The patient had just come from the funeral of a male cousin. When he told the doctor this, the doctor said, "Oh, I'm sorry, were you close to him?" The patient replied, "About 5 feet."

In short, B-C's empathizing is exactly what we already knew individuals who fall within the autism spectrum don't do very well! So imagine the surprise B-C and his colleagues must have felt when they learned that Asperger patients performed worse on the faux pas task than non-autistic individuals! I don't think I need to point out to you that if they hadn't performed worse, THEY WOULDN'T BE AUTISTIC! That's not a prediction, it's a truism. It might be somewhat surprising that they found that on a folk mechanics tasks, in which the participants had to had to predict the movement of objects in diagrams, the Asperger patients performed better on average than the non-autistic females (there was no difference between the Asperger patients and non-autistic males). From this, B-C concludes that Asperger patients do, in fact, have extremely male brains (though it would appear that at the top end of the spectrum, Asperger patients just have normal male brains).

OK, now that I've gotten my utter dismay at that totally inane experiment out of my system, I can move on to B-C's more interesting work. This work is on the actual relationship between theory of mind and autism. He and his colleagues have conducted more than 30 experiments, most of which used some version of the false belief task, on children with autism, and in almost every case, autistic children show developmental delays in theory of mind. Of course, there are problems with the use of false belief tasks in testing theory of mind abilities. He and his colleagues have also conducted neuroimaging research, focusing on the amygdala. The amygdala is thought to be important for social reasoning, particularly in the recognition of thought, intention, and emotion in the faces of other individuals. In one experiment, individuals with Asperger syndrome and normal individuals (matched for IQ, again) looked at photos of people's eyes (which also show the eyebrows, and parts of the nose and forehead), and attempted to guess what the people were thinking. Non-autistic individuals perform quite well on this task, while Asperger patients perform very poorly. While the participants were performing the task, their brains were scanned using an fMRI machine. Consistent with the theory of mind theory of autism, non-autistic participants displayed increased activation in the amygdala, while Asperger participants displayed no increased activation, when performing the task4. If you want to look at the figure above (from the Baron-Cohen et al. paper, see footnote 4), the sides are reversed (the right side of the figure is the left side of the brain), the yellow indicates that the area was only active in the non-autistic participants, the red indicates that the area was only active in the Asperger participants, and the blue indicates that the area was active in both. The left amygdala is that little yellow area of activation that is in the middle (from top to bottom) and slightly to the right (meaning it's on the left side of the brain) in the image labeled -7.

In addition to the amygdala theory of autism and theory of mind, others have argued, speculatively, that a type of neuron that may be found in the human left prefrontal cortex (they're found in the analogous region in monkeys, but there's no direct evidence, as of yet, that they exist in humans) called mirror neurons may be involved in autism as well. In monkeys, mirror neurons facilitate imitation. In humans, left frontal cortex lesions are associated with deficits in the ability to imitate the actions of others. Autism patients also tend to show lower levels of activity in the left medial frontal cortex than non-autistic individuals (e.g., in the Baron-Cohen et al study I just described) in response to theory of mind tasks. Some have speculated, then, that autistic individuals may not have a fully functioning mirror neuron system5. If there is a mirror neuron system in the human left medial frontal cortex analogous to the one in monkeys, then it is likely that this system is associated with theory of mind, and lower levels of activation in this area in autistic individuals indicates that, mirror neuron system or no, an area of their the brains associated with imitation is not as active as it is in non-autistic individuals.

There are some problems with the theory of mind theory of autism. For one, some autistic patients do perform well on theory of mind tasks, and by adolescence, most are able to perform the most common theory of mind task, the false belief test. In addition, it doesn't seem to explain the range of symptoms in the autism spectrum. Asperger patients clearly have serious theory of mind deficits, but show few of the other cognitive or linguistic deficits that characterize more extreme forms of autism. If theory of mind deficits are the primary symptom of autism, how can this be? It certainly doesn't explain B-C's (shitty) data suggesting that autism patients are extreme systemisers. At most, the theory of mind theory just describes the systemizing-empathizing symptoms (remember, people can be balanced -- being good at one doesn't preclude being good at the other). Furthermore, neither the theory of mind theory or B-C's systemizing-empathizing distinction speaks to another aspect of autism. On some types of tasks, even highly autistic individuals tend to perform better than non-autistic individuals (I'll describe some of those tasks in a moment).

The holes in the theory of mind theory of autism have led some to look for alternative theories. The most prominent alternative theory today is the weak central coherence theory, first espoused by Uta Frith6. Frith uses "central coherence" to refer to the tendency of non-autistic individuals to process information globally, e.g., at the level of the scene as opposed to the level of individual objects in the scene. It is this tendency that is responsible, among other things, for the schematic memory effects that have been demonstrated by cognitive psychologists over the last 40 years, and of which I am so fond (as you may know, if you've been reading this blog). In schematic memory effect, people tend to remember the "gist" of a story or scene quite well, but forget many of the specific details. This is because while they encode the details, they don't spend much time processing them, instead focusing their resources on the overall context. Autistics, Frith argues, do the opposite: they focus on the details, to the exclusion of the overall context. Whereas non-autistic individuals process information globally, autistics process it locally.

What's most interesting about this theory is that it makes some very interesting predictions. For instance, it predicts that in certain tasks in which the global focus of non-autistics produces errors or poorer memory, autistics will outperform them. These predictions have been confirmed in several different types of tasks. For example, in one experiment7, Francesca Happéé had autistics look at illusion-inducing visual stimuli, such as the Titchner Circles:

These circles produce a classic perceptual illusion. Non-autistic individuals tend to say that the middle circle in the group on the right is larger than the middle circle in the group on the left, despite the fact that they are the same size.Happépé argues that this is because autistics, because they process each of the circles at the local level, are immune to the illusion-inducing effects of the overal context (the surrounding circles). Elsewhere, she has detailed several other cases in which autistics display similar "local perception biases," including a reduced susceptibility to the McGurk effect, in autistics8.

Autistics also perform better on a different sort of task that requires segmenting parts of a scene instead of analyzing the scene as a whole, or Gestalt. This task, called the Embedded Figures Task, involves finding a shape, or several shapes, that is embedded within a complex scene. Non-autistics tend to be pretty slow at this. Autistics, it turns out, are able to find the shapes faster than non-autistics9. This is consistent with the weak central coherence theory: autistics are able to find the shapes faster because they are processing local features, and are not distracted by the overall Gestalt. Further supporting this view, using the Embedded Figures Task and fMRI, Ring et al.10 found that non-autistic participants showed greater activation than autistic participants in brain areas associated with working memory in the right dorsolateral prefrontal cortex, while autistic participants showed greater activation than non-autistic controls in ventro occipitotemporal regions (the cuneus, infrior occipital gyrus, and middle occipital gyrus), which are parts of the visual system. They interpreted these results as indicating that the control participants were utilizing working memory to analyze the scene as a whole, while autistic participants were using the visual system for "object feature analysis."

In addition to areas in which autistics' performance is superior to that of non-autistics, the areas in which autistics perform as well as non-autistics support a weak central coherence theory as well. For example, while autistics are worse than non-autistics at recalling information that requires global processing (e.g., temporal sequences and the source of a memory), they are just as good at memory tasks that are not always subject to schema effects, such as recognition memory11.

But the most powerful test of a theory of autism is to use it to explain the deficits that appear in autism. This is what the theory of mind theory of autism was designed to do, and the weak central coherence theory will have to do it as well if it is going to be a viable alternative. I'll give one such explanation of a deficit here, and then describe how it might be related to some of the other deficits. A couple years ago, Blake et al.12 conducted an interesting experiment in which they presented autistic children and non-autistic controls with two types of motion. One type involved small lines moving together to form a group. The other involved point-light animation of human movement (see the figures on this webpage to get an idea of what point-light animation is). Autistics and controls performed equally well on the line movement task, but autistics were much worse at recognizing biological motion in the point-light animation task. Blake et al. interpreted this finding as support for the theory of mind theory, arguing that deficits in theory of mind make it more difficult to interpret biological motion.

But there's another possible interpretation. In the same year, Bertone et al.13 published a study in which they presented autistic and control participants with tasks involving two different kinds of motion, and asked them to indicate the direction of motion. The two types of motion in this study were first and second-order motion. Previous neuroscientific research has shown that first-order motion (in this study, motion defined by changing the luminance across the screen), which doesn't involve the integration of multiple sources of information, is processed early in the visual system in the primary visual cortex, while second-order motion (in this case, motion defined by changing texture across the screen) is processed further along in the visual system, because it requires integrating multiple sources of visual information. Bertone et al. predicted that, consistent with the weak central coherence theory, autistics would peform well on first-order motion tasks that do not involve integration, but poorly on second-order visual tasks that do, and thus demand a more global processing approach. And that's what they found. Biological motion tasks like those used in the Blake et al. study are second-order tasks, in that the viewer has to integrate the information (often conflicting, directionally, at the local level) in order to perceive and identify biological motion. It could be, then, as Bertone et al. argue, that difficulties in processing complex, integrated visual information underlie theory of mind deficits in autism, rather than the other way around. Much of what makes up theory of mind in humans involves integrating multiple sources of sensory information at a higher level (e.g., faciaexpressionsns, bodily movements, and auditory cues including verbal and nonverbal expressions). If the primary deficit in autism is one of integrating information at a global level, it stands to reason that theory of mind would suffer (perceiving biological motion certainly would).

Baron-Cohen and some others feel that the weak central coherence theory is actually a compliment to the theory of mind theory. They don't think it can stand on its own. It doesn't, for instance, explain B-C's systemizing v. empathizing data. Then again, neither does the theory of mind theory, but B-C's never let that deter him. But for all intents and purposes, the jury is still out. As of yet, there's no direct data that can directly explain the relationship between global processing difficulties and theory of mind deficits in autism. They may be causally related, they may share a common cause, or they may be completely unrelated. Neither theory can as of yet explain all there is to explain about the cognitive deficits in autism, so it's unlikely that either will end up standing alone. We'll have to wait and see.

1 Lawson, J., Baron-Cohen, S., & Wheelwright, S. (2004). Empathising ansystemizingng in adults with and without Asperger Syndrome. Journal of Autism and Developmental Disorders, 34(3), 301-310.
2 Baron-Cohen, S. (1999). The extreme male-brain theory of autism. In H. Tager-Flusberg (Ed.), Neurodevelopmental Disorders. MIT Press.
3 Lawson, Baron-Cohen, & Wheelwright (2004).
4Baron-Cohen, S., Ring, H.A., Bullmore, E.T., Wheelwright, S., Ashwina, C., & Williams, S.C.R. (2000). The amygdala theory of autism. Neuroscience and Biobehavioral Reviews, 24, 355–364. The brain scans are from Figure 2, p. 361.
5 Williams, J.H.G., Whiten, A., & Perrett, D.I. (2001). Imitation, mirror neurons and autism. Neuroscience and Biobehavioral Reviews, 25, 287-295.
6 Frith, U. (1989). Autism: Explaining the Enigma. Blackwell Science.
7 Happé, F.G.E. (1996) Studying weak central coherence at low levels: children with autism do not succumb to visual illusions: a research note. Journal of Child Psychology & Psychiatry, 37, 873–877.
8 Happé, F.G.E. (1999). Autism: Cognitive deficit or cognitive style? Trends in Cognitive Sciences, 3(6), 216-222.
9 Jolliffe, T., & Baron-Cohen, S. (1997). Are people with autism and Asperger syndrome faster than normal on the Embedded Figures Test?
Journal of Child Psycholology & Psychiatry, 38, 527‚–534.
10 Ring, H.A., Baron-Cohen, S., Wheelwright, S., Williams, S.C.R., Brammer, M., Andrew, C., & Bullmore, E.T. (1999). Cerebral correlates of preserved cognitive skills in autism: A functional MRI study of Embedded Figures Task performance. Brain, 122, 1305-1315.
11 Bemnetto, L., Pennington, B.F., & Rogers, S.J. (1996). Intact and impaired memory functions in autism. Child Development, 67(4), 1816-1835.
12 Blake, R., Turner, L.M., Smoski, M.J., Pozdol, S.L., & Stone, W.L. (2003). Visual recognition of biological motion is impaired in children with autism. Psychological Science, 14(2), 151-157.
13 Bertone, A., Mottron, L., Jelenic, P, & Faubert, J. (2003). Motion perception in autism: A "complex" issue. Journal of Cognitive Neuroscience, 15(2), 218-225.

Sunday, August 21, 2005

CogBlog: Tomasello, Chapter 1

What is unpleasant and threatens my modesty is that in fact I am every name in history.

Nietzsche wrote that in a letter to Jacob Burckhardt in January, 1889, after his descent into insanity, but if you look past the self-absorption, in a disturbing and poetic fashion the sentence manages to captures the essence of the hypothesis that Tomasello defends in The Cultural Origins of Human Cognition: our uniquely human cognitive abilities exist because our minds "stand on the shoulders of giants," or contain within them the advances of the human minds that came before us. In Chapter 1, Tomasello puts forth this hypothesis, the basic reasons he thinks it is necessary (the "problem"), and in a very general way, the themes that will guide his defense of it. As Tomasello himself notes, Chapter 1 thus serves as a précis for the rest of the book. That means that it's important that we understand the basic concepts in the chapter, and also that we can begin to see some of the potential problems with Tomasello's view. In this post, I will briefly summarize the main points in Chapter 1, and describe one of the potential problems that may threaten the entire exercise.

In the beginning, Tomasello lays out the problem that motivates his hypothesis. Six million years ago, the evolutionary paths of modern apes and modern humans diverged. Until two million years ago, the evolutionary path of modern humans was dominated by one genus, Australopithecus, which was much more ape-like than human-like, particularly in its cognitive abilities. And it appears that most of our modern cognitive abilities arose within the last 250,000 years. Tomasello believes that this time period does not allow enough time for the evolution of these abilities through the ordinary mechanisms of biological evolution. Hence the problem: how can there be such large cognitive differences between apes and humans if there was not enough time for them to evolve through the ordinary biological evolution?

If there wasn't enough time for biological evolution to produce our uniquely human cognitive abilities through its ordinary mechanisms, then another mechanism is required, and Tomasello's hypothesis is that the mechanism through which most of our uniquely human cognitive abilities developed is cumulative cultural evolution. While other animal species are able to transmit information socially, they are not able to do so cumulatively, and therefore, as Blar notes in his first post on Tomasello, their inability to build on the past leaves them "doomed to repeat it." In Chapter 1, Tomasello gives us his hypothesis for how humans evolved the ability to transmit knowledge culturally. There are two primary parts to this hypothesis, one of which is responsible for the other: the ability to share attention with concspecifics, and the resulting development of uniquely human linguistic abilities. I'll briefly discuss each.

As Tomasello notes, humans and nonhuman animals share many of their basic cognitive skills. Tomasello lists perception, memory, attention, and categorization (p. 10) as shared cognitive abilities. He also notes that nonhuman primates, particularly those closest to us evolutionarily (chimpanzees and bonobos) are capable of innovation, or the creation of new knowledge and abilities (p. 5). But unlike humans, these nonhuman animals are unable to transmit their innovations to future generations. For Tomasello, there is one difference between humans and all nonhuman animals that allows us to develop the ability to transmit information cumulatively through culture (what he calls the ratchet effect). In order to do so, they must be able to understand that other humans have beliefs and intentions (p. 7-8). This understanding allows humans to reason about the actions of others, and this in turn allows us to imitate those actions intelligently. Thus the single difference between humans and nonhuman animals that evolved about a quarter of a million years ago must be one that makes this understanding possible. Tomasello's hypothesis is that this single evolved difference is the ability to identify with others of our species. This allows us to engage in "joint attentional activities," in which we recognize that another is attending to a specific object, and that he or she has beliefs and goals (intentions) that relate his or her actions to that object. This in turn allows us to imitate and learn from those actions. As we will see in later chapters, it also allows us to directly teach our knowledge and skills to others. In other words, once we are able to identify with other members of our species and engage in joint attentional activities with them, we are able to teach and learn collaboratively. According to Tomasello, it is these abilities that "form the basis for children's initial entry into the world of culture" (p. 7-8). Collaborative learning raises the human mind above the minds all of its evolutionary predecessors and all current nonhuman animals.

To illustrate this, Tomasello describes two examples that are missing one of two pieces of this puzzle. In the first example, autistic children are embedded in the "world of culture," but cannot enter it, because they lack the theory of mind abilities that allow us to identify with our conspecifics and thereby reason about their beliefs and intentions. This is why extremely autistic children display such profound deficits in uniquely human cognitive abilities, particularly language. In the second example, a child has the ability to identify with other humans, and thus to learn collaboratively, but grows up in an environment in which she is completely isolated from all other humans, and thus cannot benefit from the collaborative learning of accumulated cultural knowledge. This hypothetical child, Tomasello believes, will resemble the extremely autistic child in its cognitive abilities.

It's here that one objection might be raised against Tomasello's hypothesis (Clark raises a similar one in his post, and Jesse discusses it as well). The idea that the primary deficit from which autistics suffer is in the area of theory of mind (the ability to identify with conspecifics) is still somewhat controversial. A great deal of research has been conducted on the relationship between theory of mind and autism over the last 10-12 years, but most of it has relied on the false belief task (see the previous post on Tomasello for a description of this task). However, several researchers have argued that the false belief task does not actually test for theory of mind. Bloom and German, for example, have argued that the false belief task is a test of more than just theory of mind, and that there is more to theory of mind than is tested in the false belief task. Thus, individuals may have theory of mind abilities, but have deficits on other abilities that are tested in the false belief task, and thus be unable to perform the task. Conversely, they may not have fully developed theory of mind abilities, but pass the false belief task because it does not test for all of the aspects of theory of mind. Thus, it's difficult to conclude from autistic children's inability to perform the false belief task that those children do not have theory of mind abilities. Furthermore, autistic children are able to perform on other theory of mind tests, albeit at a later age than non-autistic children. It may be that autistic children have deficits in other abilities that underlie theory of mind as well as many of the other cognitive deficits that autistic children display. The leading theory is still that autism is related to theory of mind deficits, but exactly what this relationship is, and its effect on the many symptoms of autism is still unknown, and the topic of a great deal of current research. If it turns out that there is a common underlying cause of theory of mind deficits and other cognitive deficits in autism, Tomasello will have lost one of the strongest pieces of evidence for his hypothesis.

Returning to Chapter 1, immediately after the discussion of the ability to identify with conspecifics, Tomasello discusses one of the most important results of this ability: the ability to learn and use a uniquely human language (p. 8-11). It is here that Tomasello writes one of the most interesting and important passages in the entire book, so I will quote it at length:
As the child masters the linguistic symbols of her culture she thereby acquires the ability to adopt multiple perspectives simultaneously on one and the same perceptual situation. As perspectively based cognitive representations, then, linguistic symbols are based not on the recording of direct sensory or motor experiences, as are the cognitive representations of other species and human infants, but rather on the ways in which individuals choose to construe things out of a number of other ways they might have construed them, as embodied in the other available linguistic symbols that they might have chosen, but did not. Liguistic symbols thus free human cognition from the immediate perceptual situation not only by enabling reference to things outside this situation ("displacement" Hockett, 1960), but rather by enabling multiple simultaneous representations of each and every, indeed all possible, perceptual situations. (p. 9, emphasis mine)
The point here is that language enhances the abilities to learn and transmit cultural knowledge, the very abilities that make the learning and use of language possible. Language opens up all sorts of cognitive avenues. Tomasello lists three of these (p. 10): metacognition, in which we think about our own thoughts and thought processes; representational rediscription, which allows us to rerepresent specific knowledge in a way that makes it possible to use it in a broader array of situations; and dialogical thinking, in which we are able to adopt multiple perspectives at once. In addition, language provides us with our most efficient method of cultural transmission. In short, then, through the creation of language, and by making it possible for children to learn language ontogenetically (i.e. developmentally), the abilities that underlie collaborative learning have in essence created an amplifier through which they can develop the full range of human culture and uniquely human cognitive abilities.

This account of the role of language fits nicely with the agent-based model of the creation of a lexicon that I described in a previous post. In simulations using that model, simple agents were able to develop a lexicon for describing all of the possible perceptual inputs in a simple environment. Furthermore, through the development of a consensus between the members of the community, the lexicon allowed the agents to better represent that environment. That consensus was developed entirely through social interactions (speaker-listeneinteractionsns). In fact, all you had to do to make it difficult for such a consensus to develop between two agents was to make it difficult for them to identify with each other. While the model is simple and idealistic, it serves as a nice existence proof of the viability and power of Tomasello's account of the role of language in amplifying cultural learning, and the power of cultural learning in developing a common language.

Now that I've summarized Tomasello's problem and hypothesis, I can briefly discuss one potential objection. Tomasello's hypothesis is motivated entirely by the existence of his problem. If it turns out that six million, or even 250,000 years, is enough time for multiple cognitive abilities to evolve through ordinary biological means, than Tomasello's hypothesis is no longer motivated. This is in fact one of the most common criticisms of the book that reviewers, particularly biologists, have mentioned. Fortunately, we have a few biologists in the group who can help us sort this issue out. In fact, one of those biologists, Bora Zivkovic (whom you may know as Coturnix) offered just this criticism in his post on Chapter 1. Bora argues that chimps may not be as similar to us as Tomasello thinks, which could mean that we must have evolved multiple cognitive abilities biologically in order to explain the uniquely human cognitive skills that we display today. Bora also notes that there are several well-known evolutionary mechanisms that could be used to explain the rapid evolution of multiple cognitive abilities. He lists the Baldwin effect, niche construction, sexual selection, and group selection (multilevel selection). Since I'm not a biologist, I won't attempt to explain these to you (read Bora's post and follow the links for good explanations), but I will note that the first two, the Baldwin effect (the link is to a PhD dissertation that provides a thorough review and critique of the use of the Baldwin effect) and niche construction, have been used in discussions of the evolution of language. If it is in fact the case that there has been enough time in our evolutionary history to develop multiple cognitive adaptations, then the rest of Tomasello's book is interesting, but the validity of the view it describes will be highly questionable.

But we haven't even begun to delve into the substance of Tomasello's arguments, so it's too early to bury his hypothesis entirely. In Chapter 2, he'll present his argument that the evolutionary time period was not sufficient for the development of multiple cognitive adaptations, and will discuss in more detail the process of "cultural inheritance." Things are just starting to get interesting.