Saturday, April 16, 2005

A Different September 11

One of my hobbies is reading about the history of psychology, and in particular, the history of psychology leading up to, and comprising, the cognitive revolution. It's not entirely for amusement, either. We can learn a lot from the historical movements of ideas in a particular science, especially in one as young as cognitive science. For example, cognitive science was born, in large part, in reaction to behaviorism, but also out of behaviorism, so in some ways it explicitly ignored some of the insights of behaviorism in order to distance itself from it, while in others it accepted some of the tenets of behaviorism unquestioningly. Even today, then, there are some largely unquestioned assumptions and biases in the majority of cognitive science research. The best example is the almost complete lack of attention to emotion and culture in cognitive scientific research. There have been movements, in recent years, to change this, especially in cognitive neuroscience, where researchers like Antonio Damasio have begun paying a lot of attention to the influence of emotion on cognitive processes, and cognitive anthropology, where people like Scott Atran, Doug Medin, and Richard Nisbett have begun to look at the role of culture in cognition. Still, the bulk of cognitive scientific research ignores these things, just as it did in the 1950s and 60s, either for philosophical or practical reasons.

While learning about important things like the inherent biases and unquestioned assumptions of cognitive science through studying its history is important, it's even more important that I learn useless trivia about my field. And one of the most important pieces of triva I've learned is the birthdate of cognitive science.

I imagine that for most sciences, there's not much of a concensus on the dates, or even the years, of their birth. We might say that modern physics was born on July 5, 1687, but couldn't we also say it was born earlier, with Kepler or Galileo? Even if we agree on the date, that's only the birthdate of modern physics (we could probably also say that modern biology was born on November 24, 1859), not the birthdate of physics (or biology) itself. But for cognitive science, there is widespread agreement about the year, and even the date of its birth. We could say that it was born in September, 1948, at the Hixon Symposium on Cerebral Mechanisms in Behavior at the California Institute of Technology, where Karl Lashley delivered the speech that likely killed behaviorism, but that date signals a death, more than a birth. It took a few years for researchers from multiple disciplines, including mathematics and computer science, psychology, neuroscience, linguistics, and philosophy, to come together and create something to replace behaviorism. And the consensus is that they finally did so on September 11, 1956.

What happened on September 11, 1956? Well, it was the second day of the Symposium on Information Theory at MIT. On that date, George Miller presented his now famous "Magical number seven plus or minus two" findings, and Noam Chomsky presented his "Three models for the description of language." Thus, Miller himself1, along with Newell and Simon2 and Gardner3 all give this date as the birthdate of cognitive science. Even if we can't agree on the exact date, the year is pretty uncontroversial. In addition to the works of Miller and Chomsky, it also saw the publication of The Study of Thinking, Claude Shannon and John McCarthy's edited book Automata Studies, and Newell and Simon's "The logic theory machine: A complex information processing system,"4 which was the most thorough description of an AI project to date. In addition, Newell and Simon, along with Marvin Minsky, McCarthy, Shannon, and Nat Rochester held a conference at Dartmouth that summer, which would turn out to be the first conference on artificial intelligence. So, in a sense, AI, cognitive psychology, and Chomskyan linguistics were all born in the span of a few months in 1956.

1 Miller, G. (2003). The cognitive revolution: A historical perspective. Trends in Cognitive Sciences, 7(3), 141-144.
2 Newell, A., and H. A. Simon. 1972. Human Problem Solving. Englewood Cliffs, N.J.: Prentice-Hall. (p. 4)
3 Gardner, H. (1985). The Mind's New Science: A History of the Cognitive Revolution, New York: Basic Books.
4 Newell, A., and H. A. Simon. 1956. The logic theory machine: A complex information processing system.
IRE Transactions on Information Theory, 2(3), 61-79.

7 comments:

Anonymous said...

G'day Chris.

Nice post, though I'm not sure I agree with it entirely. There was a nice reply to Miller's article in Trends in Cognitive Science(Houde & Mazoyer, 2003) that I think does a good job of putting things in context. The cognitive revolution of the 50s was very much an American thing. Chomsky's reponse to Skinner, and the decline of behaviourism more generally, had less impact outside the US since behaviourism was not so widely adopted across the Atlantic.

In a broader sense, one of the key features to modern cognitive science is formal descriptions of cognitive processes, which certainly goes back at least to Descartes' work in the 17th century, through people like George Boole in the 18th, even William James in the 19th. Drawing a longer bow, a lot of the ancient Greek work on deductive logic is probably best viewed as an early attempt to provide formal descriptions of human reasoning.

Even in the 20th century, I'm not sure that I see the 50s (and 1956 in particular) as quite so definitive, especially when you look outside the US. You can see a great deal of modern cognitive science in Alan Turing's work during the 40s (see: here for a nice summary), though it probably reached it's peak with his 1950 article in Mind. Even so, to quote from the Turing entry:

It is also a commonly expressed view that Artifical Intelligence ideas only occurred to pioneers in the 1950s after the success of computers in large arithmetical calculations. It is hard to see why Turing's work, which was rooted from the outset in the question of mechanising Mind, has been so much overlooked. But through his failure to publish and promote work such as that in (Turing 1948) he largely lost recognition and influence.

Basically, I reckon the "born in 1956" story is a bit of a post-hoc account made up to fit in with the creation of the corresponding academic departments. The intellectual discipline has probably been around ever since humans have had the language to ask themselves "how is it that I am thinking this?"

Chris said...

Dan,
There's certainly a sense in which the problems addressed by the cognitive sciences can be traced back to the Greeks. With Descartes, philosophy becomes even more cognitive-science like, particularly with his description of automata. Even more so with Kant, who has pretty much dominated empirical psychology and analytic philosophy in one way or another.

Then there were major works in the history of psychology and the study of mind, in and outside the U.S., which anticipate many of the ideas and themes of cognitive science. You mention James, but Ebbinghaus, Law, Bartlett, Vgotsky, Piaget, and several others could be included on the list.

And cognitive science certainly wasn't born out of a vacuum. The Hixon Symposium took place 8 years before the Dartmouth and MIT symposia, and the foundations of AI and the computational view of mind were laid during (and just prior to) WWII with the work of people like Turing and the rapid development of computing machines, and then with the publication of "The Mathematical Theory of Communication" in 1948 (not to mention Hebb's publication of the next year).

So, by the time 1956 rolled around, there was already something in the air. Strong behaviorism was all but dead (it only took Verbal Behavior and the damaging reviews to kill it completely), and people from several disciplines were starting to approach the same issues from a fairly similar perspective (the computational perspective), but it took the publication of some major works which synthesized some of the major research and ideas, most of which came in 1956, and all of which came between '55 and '60, to really launch a new discipline, and a new paradigm.

We could probably quibble about September 11, though Miller, Gardner, and Newell and Simon all agree that it was Miller's paper, along with Chomsky's, that really signaled the beginning. But when you add Newell and Simon's own book, which really took computationalism from the highly theoretical (in Turing) to a real research program, I think 1956 is a pretty good year to call the birth year. It's the year that the talk of "representations" became respectable in psychology again, that theoretical linguistics turned computational (perhaps irreversibly), and that AI really got off the ground. I haven't talked about cognitive anthropology, but 1956 was also the year that Goodenough published "Componential analysis and the study of meaning," and Loundsbury published "A semantic analysis of Pawnee kinship usage," with Conklin's "Hanunoo color categories" being published the year before. These three works pretty much launched the field of cognitive anthropology, and brought it in touch with the work of the psychologists and linguists who were important in the launching of cognitive science.

The academic departments didn't really start to appear until the 1960s and 70s. Harvard's Center for Cognitive Studies, which was really the first "cognitive science" program in the U.S., opened in 1960, and it wasn't until 1976 that "cognitive science," and even "cognitive psychology," began to be common at American universities (that's when the Alfred P. Sloan Foundation committed to putting 15 million into the field). In fact, until the 1990s, the "cognitive" program at the school where I did my graduate work was called "Human Experimental Psychology," and a cognitive science program didn't exist until 2001.

So, while you're probably right, the "birthdate" is a post hoc observation (they always are when sex isn't involved), it's one that many of the participants in the cognitive revolution agree upon (at least the year, if not the date), and it's the year that those participants, coming from many different disciplines, began to really converge ideologically and methodologically.

Anonymous said...

I don't really disagree with you on the importance of the Dartmouth and MIT symposia. I just think that it would only take a small shift in how we do cognitive science to suggest that it wasn't Miller in 1956 but Brunswik in 1943 or Solomonoff in 1963 that really changed the way we think. So, something genuinely special about 1956 ought to show up as a noticeable "jump" in frequency of papers that we'd now call cognitive science, beginning at about that time (adjusting for publication lag, obviously). I tried testing this by graphing the dates of references in my thesis, but (sans outliers) all I saw was a smoothly rising curve starting at about 1935. That doesn't prove much besides my own laziness, I'll admit, but a broader analysis across the field could be done. I'd feel safer relying on that than on the personal recollections of the same people whose ideas are currently the most influential. As far as I know, no-one asked Stephen Grossberg to provide a history of connectionism, or Chris Wallace to recount the development of algorithmic statistics.

Chris said...

Unfortunately, a similar analysis of the references in my dissertation won't do us much good. There are a few from the 17th and 18th centuries, a couple from the 20th century before 1940, and then a big gap until the mid-1960s when people started revisiting Bartlett. So, my curve would make it look like cognitive science was born in 1967.

A history of connectionism would be pretty interesting, and would go back to about the same time, but would have some big gaps while the exclusive-or problem simmered and the Minsky and Papert critique of PERCEPTRON dissuaded potential modelers. Still, it would sort of begin around the same time (late 40s to mid 50s). I wonder if anyone's ever written a thorough one, rather than a brief one in the introduction to their book. If you know of such a history, let me know. I'd love to read it (like I said, this is a hobby).

The 1956 date also ignores the history of "the ecological approach," which while it didn't take off until the 1970s, was really born in 1950 with the publication of The Perception of the Visual World, which is one of my favorite books of all time (making me somewhat of a heretic where I come from).

Still, I think it's safe to say that until the early 1980s, cognitive science and computationalism, where symbolic representation was at the center, were pretty much synonymous, and computationalism really took off between '55 and '60.

I definitely see your point, and of course, trying to pin down the date is somewhat silly (the post was meant to be silly, of course), but I'm going to stick with the latter half of the 1950s, and use 1956 (and September 11) as a sort of symbolic period. It just feels to me like the time when a new mindset became pervasive, rather than the rebellious direction of a few iconoclasts.

Anonymous said...

I'll agree with that. I think I'm just being grumpy (not sleeping well on this side of the planet). It does make me think, however, that I ought to read more of the history to the discipline. I've only read about a third of those references you talked about. It's even more depressing looking at how many of those "100-top works in cog sci" that I've not read. Oh well.

Chris said...

Dan, I went through grad school without reading many of the top 100, and if not for a course I took on the history of psychology, I wouldn't have read many of the papers I cited here (I also spent some time studying cognitive anthropology on the side, because I find it fascinating). It's only over the last 6 or 9 months that I've really gotten into the history of the discipline, and some of the historical texts. It's taken me to some weird places (so to speak, too). F.A. Hyeck's The Sensory Order, which is strikingly similar to Hebb's work, for instance. Some of the works have been a real treat. Bruner, Goodnow, and Austin's The Study of Thinking is one of the best books I've read in the field.

Which reminds me, you mentioned Brunswik in one of your comments. I have been searching all over for a good copy (preferably from the 1952 printing) of The Conceptual Framework of Psychology. I ordered one on Amazon, but it was literally falling apart when it got here. You wouldn't happen to know a good online source for old and rare books like that, would you?

Chris said...

Dan, I went through grad school without reading many of the top 100, and if not for a course I took on the history of psychology, I wouldn't have read many of the papers I cited here (I also spent some time studying cognitive anthropology on the side, because I find it fascinating). It's only over the last 6 or 9 months that I've really gotten into the history of the discipline, and some of the historical texts. It's taken me to some weird places (so to speak, too). F.A. Hyeck's The Sensory Order, which is strikingly similar to Hebb's work, for instance. Some of the works have been a real treat. Bruner, Goodnow, and Austin's The Study of Thinking is one of the best books I've read in the field.

Which reminds me, you mentioned Brunswik in one of your comments. I have been searching all over for a good copy (preferably from the 1952 printing) of The Conceptual Framework of Psychology. I ordered one on Amazon, but it was literally falling apart when it got here. You wouldn't happen to know a good online source for old and rare books like that, would you?