by Will
I should have realized that my area of interest is John's area of academic expertise. I happen to have a congenital predisposition to jumping in over my head however (confirmed by the latest fMRI scan), so with the full expectation of being corrected by somebody who knows more about the subject, let's get down and dirty with modularity.
John has gotten the argument going the right way by establishing the useful distinction between "soft modularity" and "hard modularity." For those new to the debate and unfamiliar with the terms, the humble PC provides nice examples of each:
1) I'm currently typing this post on a system running Linux. One of the first major differences between Windows and Linux which new users notice is that unlike in Windows, where installing an update to the operating system means rebooting the whole computer; the software components in Linux are much more rigidly compartmentalized, so I can apply security patches to my desktop environment without rebooting my web server. Essentially, everything is a separate program, running in a separate process on my CPU, and the different programs communicate with each other using various sorts of signals in a common language.
This is an example of soft modularity -- division at the conceptual or programmatic level which leads to independence (including, importantly, time independence) at the systems level, while all residing on the same hardware framework. Translating this idea to the brain; soft-modularity means that we think specific mental tasks (face recognition, balancing while walking, and judging the distance to objects in the universe are usual suspects) are compartmentalized into (mostly unconscious) subcomponents of our cognitive system.
2) Those of you who play lots of video games have probably at one time or another purchased an outlandishly overpriced video card. The point of a video card is that it takes an already-softly-modular set of tasks (rendering cool explosions), and offloads them onto a dedicated piece of hardware that's optimized for doing the kinds of things you need to do to render cool explosions (primarily vector operations).
The important point is that not only are the tasks conceptually and systematically independent, but they're actually taking place in a different piece of physical space -- hard modularity at work. Whenever some idiot gets published in Nature announcing that he's "found the love center of the brain", he's invoking hard modularity.
Now, if I'm not grossly misreading him, John is suggesting that it's the evidence for hard modularity that's lacking, but that soft modularity is not a big deal. My argument is twofold: evidence for hard modularity does exist, albeit not in the form of fMRI studies and not the sort which justifies some of the absurd conclusions that neuroscientists reach. At the same time, soft modularity is the much bigger deal.
On the first score, let's remember that Brodmann's now-discredited areas were originally laid out on the basis of differences in neuron-shape. To a techie like myself, that immediately smells like hardware optimization. Roughly speaking, there is no "average" neuron. Depending on the region of the brain, they vary wildly in size, shape, number of incoming connections, number of outgoing connections, etc. It doesn't seem unreasonable to suspect that these differences arise because frequently-used cognitive functions are being "hardcoded". It's true that most attempts to map suspected functional modules to specific regions have met with total failure, but that might just be because we're not smart enough to grasp the correct functional divisions. My feeling is that there's not enough evidence for either side to be gloating just yet, but common sense sugests that the modularists are right.
On the flip side, I'm actually really curious as to why John thinks that hard-modularity is the bigger deal. The way I see it, soft-modularity tells us that the brain is (or perhaps "behaves like" is more appropriate) a computer. Hard-modularity only tells us that the computer is intelligible. The latter may have greater implications for public policy, but the former is the one that smacks down people who want to believe that the mind is an undifferentiated blob of thinking stuff.
I should have realized that my area of interest is John's area of academic expertise. I happen to have a congenital predisposition to jumping in over my head however (confirmed by the latest fMRI scan), so with the full expectation of being corrected by somebody who knows more about the subject, let's get down and dirty with modularity.
John has gotten the argument going the right way by establishing the useful distinction between "soft modularity" and "hard modularity." For those new to the debate and unfamiliar with the terms, the humble PC provides nice examples of each:
1) I'm currently typing this post on a system running Linux. One of the first major differences between Windows and Linux which new users notice is that unlike in Windows, where installing an update to the operating system means rebooting the whole computer; the software components in Linux are much more rigidly compartmentalized, so I can apply security patches to my desktop environment without rebooting my web server. Essentially, everything is a separate program, running in a separate process on my CPU, and the different programs communicate with each other using various sorts of signals in a common language.
This is an example of soft modularity -- division at the conceptual or programmatic level which leads to independence (including, importantly, time independence) at the systems level, while all residing on the same hardware framework. Translating this idea to the brain; soft-modularity means that we think specific mental tasks (face recognition, balancing while walking, and judging the distance to objects in the universe are usual suspects) are compartmentalized into (mostly unconscious) subcomponents of our cognitive system.
2) Those of you who play lots of video games have probably at one time or another purchased an outlandishly overpriced video card. The point of a video card is that it takes an already-softly-modular set of tasks (rendering cool explosions), and offloads them onto a dedicated piece of hardware that's optimized for doing the kinds of things you need to do to render cool explosions (primarily vector operations).
The important point is that not only are the tasks conceptually and systematically independent, but they're actually taking place in a different piece of physical space -- hard modularity at work. Whenever some idiot gets published in Nature announcing that he's "found the love center of the brain", he's invoking hard modularity.
Now, if I'm not grossly misreading him, John is suggesting that it's the evidence for hard modularity that's lacking, but that soft modularity is not a big deal. My argument is twofold: evidence for hard modularity does exist, albeit not in the form of fMRI studies and not the sort which justifies some of the absurd conclusions that neuroscientists reach. At the same time, soft modularity is the much bigger deal.
On the first score, let's remember that Brodmann's now-discredited areas were originally laid out on the basis of differences in neuron-shape. To a techie like myself, that immediately smells like hardware optimization. Roughly speaking, there is no "average" neuron. Depending on the region of the brain, they vary wildly in size, shape, number of incoming connections, number of outgoing connections, etc. It doesn't seem unreasonable to suspect that these differences arise because frequently-used cognitive functions are being "hardcoded". It's true that most attempts to map suspected functional modules to specific regions have met with total failure, but that might just be because we're not smart enough to grasp the correct functional divisions. My feeling is that there's not enough evidence for either side to be gloating just yet, but common sense sugests that the modularists are right.
On the flip side, I'm actually really curious as to why John thinks that hard-modularity is the bigger deal. The way I see it, soft-modularity tells us that the brain is (or perhaps "behaves like" is more appropriate) a computer. Hard-modularity only tells us that the computer is intelligible. The latter may have greater implications for public policy, but the former is the one that smacks down people who want to believe that the mind is an undifferentiated blob of thinking stuff.
In fact I don't know how you would build any sensible model of the mind supposing many global processes, even assuming massive modularity. I hear a police siren, my hearing module tells me the pitch is getting lower, my memory module(s) dig up the Doppler Effect, and then some global process decides that the police car is moving away from me, and not that the siren really is getting lower. And for want of a better term, the global process that decides my memory wins out over my hearing... well, that's *me*.
Posted by: Salemicus | June 03, 2008 at 03:35 PM