Friends of Semantic Compositions

January 2009

Sun Mon Tue Wed Thu Fri Sat
        1 2 3
4 5 6 7 8 9 10
11 12 13 14 15 16 17
18 19 20 21 22 23 24
25 26 27 28 29 30 31

Site Statistics

Blog powered by Typepad

« Maybe try thinking of a donkey | Main | How not to test a hypothesis »

September 30, 2004



Nice post, Anti-Chris. ;)

I just want to make one quick comment, which is really just a lead up to reference. I think Lakoff's view of concepts, and metaphorical concepts in particular, is interesting, but ultimately fairly worthless, both scientifically and practically. With the exception of Raymond Gibbs' work, there has been very little quality empirical research generated, in over 20 years, by the conceptual metaphor theory. The one area where there has been any real success for the conceptual metaphor theory is in time concepts, where Lera Boroditsky and some of her colleagues and students have done some cool stuff, even if it's still hard to know what to make of it.

And now the reference. For a thorough survey of concept research, which will show how Lakoff's ICMs, and particular his prototype theory (which might have been current in 1975, but not in 2004), simply fail to describe human knowledge representation, anyone interested in this sort of thing should check out Greg Murphy's The Big Book of Concepts.

If, after that, you haven't become so bored with concept research that you never want to hear or read the word "exemplar" again, and you're still really really interested in understanding why conceptual metaphor theory fails to account for every important aspect of concepts, from categorization to inference to comparison, analogy, and (ironically), metaphor, there's a huge body of research on prototype theories, their competitors, exemplar theories, and a whole host of alternative theories, much of which can be found in the references of Murphy's book. If you really like Lakoff, realize he's wrong, but don't want to stray too far, Larry Barsalou's perceptual symbol systems approach (which is also detailed in Furnishing the Mind by Jesse Prinz, for those who want the philosopher's perspective), the work of Arthur Glenberg, or even Gilles Fauconnier's work. Barsalou's work is still full of holes, and with every revision, it gets further from anything resembling Lakoff's view, but it's still closer that most computational theories, and has something like ICMs within it. I personally like Fauconnier's work, because it's compatible with just about any theory of concepts (prototype, exemplar, theory theory, role-governoed, multipe system, or whatever), and because he has the best examples ("If Clinton were the Titanic, the iceberg would have sunk,"). Glenberg does very interesting research, but the conclusions he draws from it tend to be highly speculative and unwarrented given his data.

One more bit of irony that I can't resist. Lakoff's criticism of "objectivism" in philosophy targets the idea of Natural Kinds. It's interesting, then, that this paper criticizes theories like Lakoff's (though it doesn't mention Lakoff specifically) for treating concepts as Natural Kinds. It also has a short section on why prototype theories don't work.

Semantic Compositions

Much appreciated.

As I've written before, I tend to take Lakoff's work more metaphorically than literally, since I've never accepted many of the conclusions he draws from it. And while I don't want to be unduly harsh, I had my fling with "Cognitive Grammar" and found it wanting. I find Lakoff's way of laying out the theory to be a very useful tool for talking about categorization, but I'm certainly prepared to dump him in favor of something that better accounts for real data.

I don't claim to be current by any stretch in the field of conceptual metaphors. I've done a lot of work in the last 3-4 years on knowledge representation, but from the symbolic natural language processing approach, which tends to be long on folk theories of properties and categorization, and short on psychological reality (as you appear to note yourself). I'll definitely have a look into Murphy's work.


I admire, and even envy Lakoff's ability to discuss very complex subjects readably and even engagingly. Even Women, Fire, and Dangerous Things and Philosophy in the Flesh, which are thorough presentations of a fairly broad psychological and philosophical perspective are easily read by nonexperts. Pinker writes trade books for laypeople, and is deservingly praised, but Lakoff writes scholarly works that are just as approachable (Mark Johnson has something to do with this, as well). His ability to describe the complex in such readable language is, as far as I can tell, the reason his talk of framing has caught on among liberal intellectuals. It's also why I think that his scholarly works are good reads for people just getting into knowledge representation. The first two chapters of WFDT are excellent.

It's also why I agree with your disagreements about his political ideas (you're not really the anti-Chris, you see). I hope people take the basic idea -- the value of framing -- and do away with all the metaphor stuff. Sure, it's fun to read, but when you actually try to apply it to real world situations, you run into a problem: it's not the way peoples' minds work, so it's not the way liberals and conservatives really think. I think discerning readers tend to realize that, though. Anyone who thinks that they can capture entire world-views with such a simple framework is probably mad (Lakoff's not the only cognitive linguist to display this sort of arrogance; Fauconnier and Turner did title their most recent book The Way We Think). For this reason, I tend to think that a lot of this stuff is meant rhetorically, or metaphorically as you say.

The comments to this entry are closed.