Impressions from the 2006 Society for Neuroscience Meeting
place here in
last Saturday to Wednesday. As a
“neurophilosopher” I felt I couldn’t pass up this opportunity to attend, so I
paid my $145 for membership in SfN so I could get the discounted registration
price of $220 to attend the meeting (and all I got was a cheap bag)! Here are a few impressions:
1) It’s
HUGE. Imagine the biggest APA Eastern
meeting you can. Now multiply it by
10. There were over 30,000 people at SfN
(and apparently this was a low turnout year).
The convention center room where they display posters is about 3
football fields by 1 football field (in the middle, hundreds of exhibitors were
selling lots of cool and expensive equipment).
There were thousands of posters in each of the 10 poster sessions,
concurrent with about 10 sessions full of 15-minute talks and some longer featured
lectures.
2) What
could so many people have to say about the brain? Almost all of it was “low-level” data—i.e.,
about how neurons work (lots of talk about chemicals), neural connections,
genes, and how the brains of rats and mice (sometimes monkeys and
invertebrates) work. Most of the
discussion of human brains was about neurobiological disorders, but most of
that work is focused on neurochemicals, genes, and mice/rats. So, the general approach is pretty
reductionistic. I’d estimate less than
5% was about human cognition (memory, emotion, reasoning, etc.) using human
subjects. Of course, almost all of these
studies use fMRI. (This still means
there were probably over 1000 posters or talks on fMRI studies of human
cognition.)
3) It
is hard to get excited (as a naturalistic philosopher interested in human
cognition) about most of these fMRI studies.
Why? Because most of them are
designed to show that when humans do X, brain regions A, B, C light up. So, for someone who believes that doing any
cognitive task will supervene on certain brain processes, this news is not very
illuminating, especially when most of the studies just replicate previous work
(e.g., lots of interesting psychological tasks involving deductive and
analogical reasoning, all showing that BA 10—an area in the front of the
frontal cortex—lights up). The data is interesting when it shows some
interesting “cross-over” effect (e.g., that areas associated with disgust are
active during certain types of moral reasoning or that motor cortex is active
when we think about performing an action or when we watch others perform an
action) or when it tries to offer a theoretical framework for why certain parts
of the brain are involved in certain functions (though, from what I saw, this
is rare; and there was very little discussion of evolution). The upshot is that it is hard to get really
interesting data. I say these mildly
disparaging things as I embark on an fMRI study with a psychologist here at GSU
(though we are trying to get some good “cross-over” effects and develop a
theoretical framework), so I’m not saying this work should not be done. But perhaps we philosophers should feel
pretty lucky that we can pick and choose the most interesting results from the
neurosciences without having to get our hands (too) dirty, and we get to
synthesize this information and apply it to the really big questions.
P.S. I think this post got buried, but it’s worth checking
out
lobsided brain.
Eddy,
It might be too quick to dismiss fMRi work that “shows what’s already been shown,” since it is common enough in science to try to check a given hypothesis with multiple methods, e.g. paleomagnetic stratigraphy and radiometric dating for measuring the age of fossils.
One hope, roughly speaking, is that the different methods will rely on distinct assumptions that might still yield the same results. This supports the methods themselves and the conclusions. So, if fMRI results agree with other test results, that provides reason to think that maybe the many, many assumptions that go into an fMRI study are not erroneous.
Ken, I certainly wasn’t meaning to dismiss fMRI work (after all, I’m doing some!). Nor did I mean to suggest, in saying that many of the talks and posters I saw “show what’s already been shown,” that they are confirming with fMRI what’s been shown with different methodologies (that is certainly being done as well). Rather, most I saw were confirming using fMRI what’s been shown with fMRI using some slightly different task (e.g., this analogical reasoning task showed activation in BA10 and so does our analogical reasoning task).
As you suggest, we certainly need lots more work to be done to show that the assumptions that go into fMRI studies are not erroneous.
The last sentence of the post seems to me an acurrate description of the philosphers´ job, but with a caveat. It´s time for philosophers not only to collaborate with neuroscientists but to do neuroscience(or in other words neurophilosphy) with all the consequences (using neuroimaging and ohter imaginable technology) and get muddy targeting the BIG questions.
I envy you (in the good sense of the term) for being there at SFN. I´ll do my best to go to San Diego.
The triangulation or concurrency of diverse procedures and methods to prove consistency in data mentioned by Ken, have many relations to Wilson´s post regarding Coltheart´s ciritque of fMRI.
Another theme, my sincere congratulations to Gualtiero Piccinini and collaborators, even Professor Chalmers post here!
I enjoyed SFN quite a bit. Lots of good systems neuroscience and extension of imaging techniques to the in vivo preparation.
I consider myself lucky to be a neuroscientist right now: I find it pretty easy to keep up with developments in philosophy of mind (they are pretty infrequent compared to neuroscience), and I don’t have to waste time teaching undergrads about Aquinas and Sartre. :O
Seriously, though, while I understand the drive to give a neural account of ‘cognitive’ level phenomena, such theories are so damned speculative right now that most neuroscientists classify them as philosophy (I would classify them as worse than philosophy). Since the best philosophy comes after the science and (to risk sounding like a ninny) armchair philosphers will contribute little to nothing to our understanding of how brains work, I hope that more naturalistic philosophers get their hands dirty with the data, or at least with modelling (and I don’t mean the Chomsky Heirarchy: most computational neuroscientists don’t even know what it is and they are better off for it).
BTW, Atlanta is a beautiful city!
Regarding the call to arms for philosophers to roll up their sleeves and help collect data:
Do we really need more people collecting data? As Eddy observes, there are thousands and thousands on the job already. There’s a serious danger of the field and its various subfields being data rich and theory poor.
Maybe more people should roll their sleeves back down and do some serious thinking about what the point of this all is.
Eddy,
But, there is still a scientific place, I think, for confirming using fMRI what’s been shown with fMRI using some slightly different task. One wants to show that particular results are robust. (I understand that there are now hundreds of variations on the false-belief task, for example.) Checking that results are robust seems to me to be the kind of thing Kuhn would have said forms part of “normal science.” I would guess you are exactly right insofar as you are saying that most of this fMRI work does not involve and Nobel Prize winning kind of innovation and maybe even so very much innovation at all. But, then again, so much of science is not all that innovative as it is thorough.
Another thing you might be driving at is that much of the regular work with fMRI that you saw is not of that much interest to philosophers of mind. That might well be correct. [And much of the regular work of philosophers of mind is probably not of much interest to those working with fMRI.]
As an aside, William Uttal, in his book-length critique of neuroimaging, *The New
Phrenology* (or whatever), complained about the lack of replicability.
Now, it sounds as though you are complaining about too much
replication. Now I’m getting the sense, though, that that is not what you are aboud.
Like Pete, I think there is some value in philosophers not rolling up ones sleeves to do science. Much scientific experimentation involves the mastery of a lot of technical abilities, e.g. running the computer programs that convert the raw fMRI data into statistically significant results, making stimuli (e.g. images of faces) that meet certain desired specifications, learning to remove and section brain tissue, learning when to use which statistical tests, etc., etc. Mastery of these kinds of skills is an essential part of science. Philosophers of science might well have to worry about this, but maybe there is room for philosophers of mind who do not know this kind of thing. After all, once a scientist spends the better part of a decade learning to “use a hammer”, then lots of things might begin to look like nails, as they say.
I just had a long conversation with a neuropharmacology neuroscientist and a developmental psychologist about natural language. My sense was that neither one really cared that much about the poverty of the stimulus argument. It’s not so much that they thought it was wrong (which they did), but they also did not seem to care about it. Maybe this was, in part, a product of their training. To do the things each of these groups do, they just don’t have that much time or interest or inclination to learn about this argument that Chomskyan linguists put forth. This suggests something that philosophers might do. While philosophers philosophers of mind might “cut corners” by not learning about the scientific tools (e.g., statistical tests, data analysis software for fMRI, etc., etc.), we might spend our time learning about the products that have been developed from those tools, e.g. the multiple theories and approaches in cognitive science.
So, maybe there is room for trying merely reading lots and lots of diverse scientific literature very thoroughly, rather trying to contribute to the scientific literature.
I suscribe many of your words Eric. But never forget the noble goal of philosophy and even that all sciences come from philosophy in historical terms. I patiently wait Nicolelis´ book entitled “Beyond Maps” and when i learn and learn and learn… more neuroscience being capable of discriminate in wich aspects you contribute.
You are not a neuscientist, you are a philosopher. Whenever someone trys to understand how something works, what is its nature, where it come from, how it relate to to other things and finally how to get together the diverse data to make sense of it etc. that´s Philosophy with capital P. Administratively, socially and scholarly, perhaps is neuroscience or physics or chemistry or mathematics… but in essence is the noble task of doing Philosophy.
Mandik,
what is better than a philosopher collecting their own data and developing their own theories to be tested for falsifiability.
Philosophers for centuries have been producing the advancement of intellectual ideas without coat labs, in their armchairs and still are doing it amazingly.
Imaging for instance if they, dare to wear the coat lab!
What is special in philosphers is that they have by qualification, as a default condition, the capability of cocneptual construction, holistic reasoning… fit to that also labs skills. Is like having 2 in 1 (a scientist and a theoretical person in one subject)
Unfortunately, philosophy training is not very helpful for thinking about data or coming up with precise theories of how brains work. Many philosophers I talk to think that theoretical neuroscience is just philosophical neuroscience, but there are really two groups of people that don’t collect their own data. The theoretical neuroscientists (Sejnowski, Abbott, Hopfield, etc) who are trained in lots of mathematics (typically they come from physics) on one hand, and the philosophers who typically use natural language to think about brains. I think the philosophical branch of the armchair neuroscientists have done, and will do, very little to push neuroscience in fruitful directions. The mathematical branch of the armchair dwellers, though, will continue to bear fruit.
While it is possible, I just don’t see neuroscience becoming data rich and theory poor: theoretical neuroscience is exploding, especially as theoretical physicists are realizing that it is much easier to find jobs in biophysics than string theory. Theories to explain given data are a dime a dozen. While the amount of data is quite daunting, if you ask an experimentalist for their speculations about their data, you typically won’t find any shortage. However, they will tend to be quite cautious, journal editors tend to cut such speculations out of papers, and experimentalists don’t want to come off as mushy theorists in their presentation of data. There is a strong selection effect to make it look like there lots of theories. Also, in practice, it is typically experimentalists who come up with predictions that can actually be tested: this is very hard to do even if you are an experimentalist with an understanding of the nuances of the techniques.
Also, while I don’t think philosophical naturalists should necessarily be doing experiments, as I mentioned above, they would be better served by learning more mathematics and actually analyzing some data. I think Chris Eliasmith is working in this vein, and Rick Grush to some degree.
Because of such concerns, I switched from philosophy into neuroscience in graduate school, and never doubted that, given my goal of understanding how brains work, this was the best choice. To explain why I think this was the right choice would take a lot of space, but part of it is that I am continually steeped in neuroscience. I don’t waste time writing or teaching about Plato, but sodium channels and the like. When I think about brains, I don’t feel like I am deviating from my professional obligations, and I don’t waste time explaining why I am studying neuroscience. I just study neuroscience.
On the other hand, most neuroscientists don’t make far-leaping speculative links to other subdisciplines (e.g., between the molecular basis of cortical development and Chomskian linguistic theory). I think their restraint is quite justified, but it might be that having a group of people with conceptual competence in multiple yields might yield an amazing discovery.
I said:
There is a strong selection effect to make it look like there lots of theories.
When I meant, and thought I wrote:
There is a strong selection effect to make it look like there ARE NOT lots of theories.
From the armchair, the sort of thing you are describing sounds terrific. In the field, looking at actual examples, people that combine the two skill sets are dwarfed by those who don’t.
Maybe you are right (now), but i am confident taht in the future if we want to put philosophy ahead into the 21st century, we need that kind of person.
Thanks for the clarification, Eric. I was scratching my head wondering what the point was you were trying to make.
I think, overall, you’ve made a very nice comment.
I have read your post many times. You say something really important here:
“The theoretical neuroscientists (Sejnowski, Abbott, Hopfield, etc) who are trained in lots of mathematics (typically they come from physics) on one hand, and the philosophers who typically use natural language to think about brains. I think the philosophical branch of the armchair neuroscientists have done, and will do, very little to push neuroscience in fruitful directions. The mathematical branch of the armchair dwellers, though, will continue to bear fruit.”
I agree. Especially with the last sentence. It seems to me that (for example) the notion of probability that is used in many current theories of theoretical neuroscience as well in many models of brains (like Friston`s (???) models of cortical responses and so on) may need a good old-fashioned philosophical analysis.
However, I´d like to add that it is not enough for a philosopher of (cognitive) neuroscience to know about the current situation of neuroscience or cognitive sciences. It would be extremely important that philosophers would understand that many of the (mathematical and other) models (like Grush`s dear Velman-stuff) are used in other diciplines (like econometry and economics) as well. It may well be that (i) other special sciences and (ii) the philosophy of them may offer some important contributions to philosophy of neurosciences as well.
Anna-Mari wrote:
It seems to me that (for example) the notion of probability that is used in many current theories of theoretical neuroscience as well in many models of brains (like Friston`s (???) models of cortical responses and so on) may need a good old-fashioned philosophical analysis.
While I think philosophy of probability is interesting (e.g., all the fights amongst frequentist, Bayesians, and others), it doesn’t seem that probability in neuroscience is any different from that in physics. Scientifically, it is most fruitful to just use probability to model something that is not deterministic (or to do good statistics). Worrying about what probability is would slow me down.
That said, many neuroscientists don’t understand the mathematics they are using. For instance, they blithely use Shannon’s information theory, assuming it measures things that it doesn’t actually measure (I wrote a paper on this which can be found here).
I don’t know if that paper was philosophy: I would classify it as uncovering assumptions and uses of two spheres of applied mathematics. All of the paper invokes standard probability theory, what you are allowed to infer from what, but never addresses foundational issues in probability.
Eric,
Thanks for your comment.
“While I think philosophy of probability is interesting (e.g., all the fights amongst frequentist, Bayesians, and others), it doesn’t seem that probability in neuroscience is any different from that in physics. “
Yes, philosophy of probability is interesting. Unfortunately I have to say that I do not know much about it. But I have had an impression that in many probabilistic models (in current neurosciences) the notion of probability is bayesian, i.e. subjective. I do not know enough about physics, but are they really using subjective notion of probability there too?
By the way; I am really interested in your Shannon paper. We have a “information theory club” here in Helsinki. Your paper was just added to our program
.
Anna-mari:
I think most neuroscientists who use probability theory don’t have strong opinions about foundations of probability theory, though if pressed they would probably give some kind of desultory frequentist response. They just use it as a modelling tool (just as you can use differential equations without having an opinion on Platonism versus conventionalism or whatever). While we use Bayes’ theorem, that isn’t the same as being a Bayesian (Bayes’ theorem is just that: a theorem of probability theory that anyone, of any school of probability would agree with; Bayesianism is a perspective on what probabilities actually measure).
Thanks for adding the paper to the list. Let me know if you have any questions. I wrote up a minimal probability refresher as background for the paper. People who didn’t have a strong probability background found it helpful for working through the paper.
Anna-mari:
I think most neuroscientists who use probability theory don’t have strong opinions about foundations of probability theory, though if pressed they would probably give some kind of desultory frequentist response. They just use it as a modelling tool (just as you can use differential equations without having an opinion on Platonism versus conventionalism or whatever). While we use Bayes’ theorem, that isn’t the same as being a Bayesian (Bayes’ theorem is just that: a theorem of probability theory that anyone, of any school of probability would agree with; Bayesianism is a perspective on what probabilities actually measure).
Thanks for adding the paper to the list. Let me know if you have any questions. I wrote up a minimal probability refresher as background for the paper.
Eric,
You say: “most neuroscientists… if pressed they would probably give some kind of desultory frequentist response”. Could you open this a bit, please? Why do they give that response?
( I am not sure, whether I do understand this correctly…)
***
I´ll let you know. Thanks.
My point is a trivial one: someone can use probability theory successfully in science while knowing nothing about the philosophical arguments about the nature of probability.
As for why most neuroscientists are frequentists, I’m not sure. Probably because it is what is usually taught in introductory statistics courses: the probability of something happening is the frequency with which it will occur if you do lots of identical experiments. Bayesianism sounds fuzzy-headed and is not usually presented except in passing, and they aren’t aware of other more exotic interpretations of probability (e.g., propensity interpretations).
A philosophical attempt to give an account of information using counterfactuals instead of probabilities is written by Jonathan Cohen which can be found here. Note I am NOT endorsing the paper, but thought you might be interested if you are interested in information theory from a philosophical perspective.
I just mean that most biologists are taught to think of probabilities as relative frequencies, but they aren’t particularly committed to this interpretation and don’t know anything about philosophical arguments about the foundations of probability theory. For the most part, I think this is fine.
My point is a trivial one: someone can use probability theory successfully in science while knowing nothing about the philosophical arguments about the nature of probability.
– Yes, of course. However, it may turn out to be extremely interesting to figure out what kind of commitments one is then actually doing. There are at least two interesting issues here.
One issue is this: I was yesterday at the meeting of our research group and somebody mentioned a paper that is published recently. (I cannot remember the authors just now. ) The point of the paper was something like this: Science can be seen as a collection of travelling (mathematical and other) models, i.e. same models travel trough different diciplines. I thought it was pretty interesting; I already mentioned somewhere that may be philosophers should be aware the fact that same models that are used in neurosciences are used in other fields as well.
I wrote about “Velman-stuff” in my earlier post. I, of course, meant “Kalman-filters”
. Anyway, the point is that same mathematical model is travelled from motor control theory to cybernetics, econometrics, economics, to emulation theory and so on. Perhaps something similar has happened with the models of probabilities as well.
The second issue is this: ” As for why most neuroscientists are frequentists… Probably because it is what is usually taught in introductory statistics courses: the probability of something happening is the frequency with which it will occur if you do lots of identical experiments. Bayesianism sounds fuzzy-headed and is not usually presented except in passing, and they aren’t aware of other more exotic interpretations of probability (e.g., propensity interpretations).”
– Yes. I wish I would have studied mathematics when I was in that age. Unfortunately I am now too old to learn it properly. So this is a mere speculation, but perhaps there is something interesting, if one considers the theories that are used in some parts of neuroscience.
Eric, there are couple of questions I´d like to ask you related to this issue. However, the blog may not be a right place for it. I tried to find your e-mail address, but I did not. Could you be so kind and contact me, please?
***
A philosophical attempt to give an account of information using counterfactuals instead of probabilities is written by Jonathan Cohen which can be found here. Note I am NOT endorsing the paper, but thought you might be interested if you are interested in information theory from a philosophical perspective.
– Oh yes, we are. My understanding is, unfortunately, very limited.
Sure: my email is thomson | at | neuro | dot | duke | dot | edu.