Can Cognitive Neuroscience Tell Us Anything About the Mind?
Cognitive neuroscience - essentially brain scanning - has become all the rage in psychology and related fields. Given the headlong rush by, well, practically everyone, into cognitive neuroscience I still entertain a quaintly unfashionable stance: I'm sceptical. My scepticism is not total though, many cognitive neuroscientists claim that there are many exciting findings to come. They're probably right, but while neuroimaging can certainly tell us much about the brain, there's reason to believe it hasn't told us much about the mind. To understand what I mean by this we need to go back to basics by asking what research is for.
"Does cognitive neuroscience really have the power to distinguish between psychological theories?"Like all scientists, psychologists are continually knocking out new theories to explain the way we think and behave. One of the most important functions of research is its ability to differentiate between two theories. If research doesn't at least provide a clue one way or the other then theoretically, and so practically, it's a waste of time. Does cognitive neuroscience really have the power to distinguish between psychological theories? Is it any use to a cognitive psychologist?
A critic's view
Max Coltheart is Professor of Psychology at Macquarie University in Australia and in a recent journal article he wonders whether cognitive neuroscience has really told us anything useful about the mind so far (Coltheart, 2006). It's important to realise that his emphasis is on the mind, as in cognitive processes, as distinct from the brain, as in physiological processes.
"Neuroscience's strength is in physiological processes..."There's no doubt the mind's cognitive processes are a function of the brain's physiological activity but these two things are nevertheless (currently) separate questions. Cognitive neuroscience's strength is in physiological processes, and as imaging technology improves, so will the importance of its findings in this area. But, again, why should a psychologist care that much which part of the brain lights up in a scanner, if the mind's functioning is still so far removed from our understanding of its physiology?
An example
All this can be difficult to grasp in abstract. Take one of Coltheart's examples. Suppose you're a psychologist interested in how people work out what other people are going to do. Their intentions. Suppose there are only two competing theories that you've got to choose between:
- 'Simulation theory': I literally run a crude simulation of your mental state in my own mind. From this I try and work out what you're going to do next.
- 'Theory theory': I create a theory about you, then try to work out what you're going to do from that.
A recent neuroimaging study claimed to be able to distinguish between these two theories. Ramnani and Miall (2004) put people in the brain scanner, got them to carry out certain tasks and predicted that if a particular part of the brain was activated it supported the first theory, and if another, then it supported the second. What actually happened was nowhere near this simple. Despite the claims of the study's authors, Coltheart argues that actually neither theory was substantially supported or refuted by the findings.
"...evidence from cognitive neuroscience fails to distinguish between theories."This is just two theories and one study - not exactly a scathing criticism of the whole of cognitive neuroscience. But Coltheart does run through four other examples where evidence from cognitive neuroscience fails to distinguish between theories. Again, remember that we're talking about relatively high level psychological theories here, not low-level physiological processes.
Coltheart goes on to pull quotes from a range of people who argue that, in principle, neuroimaging is useless for psychological theory and understanding of the mind. Here's a good computing metaphor:
"No amount of knowledge about the hardware of a computer will tell you anything serious about the nature of the software that the computer runs. In the same way, no facts about the activity of the brain could be used to confirm or refute some information-processing model of cognition." (Coltheart, 2004, p.22)
I personally don't know enough about cognitive neuroscience to argue whether or not this statement is true, but it certainly has intuitive appeal. Considering the enormous quantity of money going into cognitive neuroscience right now, it seems unlikely this would be a majority view amongst psychologists. Not that scientist are slaves to money, of course...Ahem...
What's your view?
There's precious little discussion of Coltheart's criticisms in the blogosphere and there's plenty of good neuroscience blogs around. Let me know what you think, leave a comment below.
References
Coltheart, M. (2004) Brain imaging, connectionism and cognitive neuropsychology. Cognitive Neuropsychology, 2, 21-25.
Coltheart, M. (2006). What has functional neuroimaging told us about the mind (so far)? Cortex, 42(3), 323-31.
Ramnani, N., & Miall, R. C. (2004). A circuit in the human brain for predicting the actions of others. Nature Neuroscience, 7, 85-90.
Join 22301 readers
Hi Jeremy,
I really enjoy your posts and I've been a subscriber of your feed for a few weeks now. However, it really bugs me that from my thirty-something feeds, yours is the only one that doesn't display the whole post. Ususally, I only get the first paragraph. Can you change it so that we can get the whole "package" through RSS? Thanks for your insightful posts and best!
Hey, thanks for pointing out the Coltheart paper. I don't read Cortex, so I might not have noticed it otherwise. First, I think you should be careful when you use the term "neuroscience" to refer largely (entirely?) to cognitive neuroscience, and neuroimaging in particular. Other areas of neuroscience, including neuropsychology that comes close to (if it doesn't qualify as) cognitive neuroscience, have taught us things. And the double dissociation method can be very powerful in testing theory.
That said, I mostly agree with the critique of neuroimagining, and I've said so many times publicly and privately. In addition to the problems with neuroimaging methods, especially the subtraction method, any imaging study will be so reliant on theory (for determining what areas to look at, and for interpreting differences in activation), that it's hard to know what it tells you. And no one has really developed an evidence-based theory of what differences in activation mean anyway.
Honestly, I'm not sure that imaging will ever tell us much. Imaging requires a commitment to a strong belief in that the localization of functions in the brain is universal. And while that may be true at some level (vision utilizes the same cortices across individuals), it may very well not be true at the level at which imaging operates.
Oh, and one thing I dislike about Coltheart's critique: his very old-school functionalism, in which he treats cognition (software) as entirely amodal. It's the strong AI belief that you can take any cognitive program and implement it in any hardware. That's one hell of a philosophical assumption.
Chris, you're welcome. And thanks for reminding me about the term neuroscience. I meant cog-neuro but I forgot to write the cog! I've changed that now.
Also happy to hear from a fellow sceptic.
Any supporters of cog-neuro out there? I've read that cognitive neuroscientist say when they've got enough studies they'll be able to tell us something useful. They argue that it's a matter of steadily accumulating data.
What exactly do you mean by "the mind"? Do you have any genuine evidence that such an entity exists? I understand what the term means in popular parlance, but is it actually meaningful in any concrete or scientific sense? Why should neuroscientists be concerned about competing theories about a phenomenon that may be merely lingusitic? I have no brief for or against cognitive neuroscience, but I am skeptical of claims about the existence of a "mind" that's separate from the brain.
Greg, thanks for your comment. I'm certainly not suggesting the 'mind' is separate in any physical or philosophical sense from the brain. To rely on that well-worn quote: "The mind is what the brain does." Unfortunately there's this huge gap between our knowledge of the brain's physiological functions and it's high-level operations, namely consciousness.
An example: one of the most important functions of the mind is memory. Psychologists have been studying this for ages and are pretty sure it exists (in most people).
This article is questioning what a cognitive neuroscientist can possibly tell you about memory. What can you understand about memory other than which parts of the brain are more active in certain situations? Does this low-level information lead to any high-level conclusions about how memory works? Can you distinguish between, for example, two theories of how memory operates?
Thank you for your clarification! I actually like your blog very much and have found much to ponder in it.
You're welcome Greg, I'm glad you enjoy it.
Also, to the very first commenter on this post 'Anonymous', I'll consider this, but I know many other feeds work in a similar way to mine.
For one thing you miss out on the comments if you don't come to the site. I know you can still comment by clicking through from your feed reader but I think it encourages more participation this way.
Jeremy,
Much work has been done learning how the brain works at the (nuts-and-bolts) level of neurons, and hundreds of years have been spent formally studying the brain from the (finished-product) level of behavior and understanding. I can't help but wonder if cog-neuro is an important first step in bridging the gap between the two. Although it may, by itself, teach us very little about the brain, it may give us important clues such as where and how to study the different parts of the brain.
Thoughts?
I would like to take issue with Coltheart's comparison with knowledge of computer hardware and the uses of functional imaging. In his summary you reproduce (unfortunately I can't get access to the full paper online, and admit making a critique based on a single paragraph is rather shaky ground), the position put forward seems to me to be an apples and oranges comparison: some (computer) hardware in the "off" state and the brain in a snapshot of its "on" state.
It is clear that simple knowledge of the hardware would provide no knowledge of the software running upon the machine. Were one to take snapshots of a computer's running states over time, however, it would be possible to build up a picture of what the computer is doing easily -- all the information you need is there, both the data and the program is stored in the computer's "on" snapshots, and you would be able to see how these progress. I would posit that this time-based snapshot of the "on" state is more similar to functional imaging. The brain, of course, may not have all its functions "on show" in the same way as a computer (which seems to be what you imply cognitive neuroscientists may expect).
As described by Coltheart's paragraph, knowledge of the computer's hardware is more analogous to expecting to discover the brain's functioning purely by dissecting a dead and inactive brain. Which I expect is impossible.
Functional imaging is great for finding brain regions that are sufficient for a particular phenomenon. However, to know it is necessary requires the out-of-fashion lesion method.
It's a pity, but there does seem to be a separation occuring in these methods. Fellows had an article a while back showing how the imaging literature is pretty much ignoring the lesion literature, wheras the lesion literature makes use of the imaging literature.
To know a particular region is necessary can be shown using lesion methods. Both imaging and lesion methods have weaknesses, if they work together, we can make much stronger inferences.
We need a more integrative approach.
Anonymous #2, you're talking about bridging the gap between the neurobiology and psychology. Yes, I think you've put your finger on why, intuitively, people feel cognitive neuroscience is/will be useful.
Mike, good point. But I still think Coltheart has a strong argument, as Chris says in an earlier comment: how do you know what activation in a particular part of the brain means? This question also speaks to what anonymous #2 is saying above. The temptation is to see cog-neuro as bridging the gap, but what evidence is there of any bridging going on?
Anonymous #3, yes and the great thing about lesioning studies is you have much stronger evidence that a particular part of the brain is really involved in a whatever function you're examining. Integration is certainly the way forward. I'm constantly amazed in all areas of psychology why there isn't more integration. Or is that like saying: why can't we just all stop fighting, be nice to each other and live in peace? It's not human nature.
There are at least two questions being asked here:
What does imaging tell us about the brain? (see Chris' comments)
What does the brain tell us about the mind?
It should be noted that the field of cognitive neuroscience is interested in both questions, and progressively takes an integrative methodological approach in which I believe brain imaging may have import. That said it is not a minority opinion that 'imaging for the sake of imaging' is a waste of time, and more generally that the 'where' question may be turn out to be largely disinteresting. In the history of science, this kind of machine driven fascination is nothing new.
The second question is certainly interesting, and the numerous attempts to progress our understanding in this regard can be traced from Sherrington's conception of 'the enchanted loom' to Fuster's most recent efforts ('Cortex and Mind' 2003).
Fuster sees a desparate need to identify linearity between the cortical and mental. The crux of Coltheart's argument would seem to be that no such linearity exists, but as Chris points out, his overt functionalism suggests that he really hasn't thought about it that much. Fuster is far less sceptical, and his favoured approach (a PDP network model of cortical representation as physiological - psychological interim, termed the 'cognit') makes me optimistic too.
But finally let's not forget that the study of the mind is similarly the prey of the sceptics - it is no easier to go from mind to brain. Therefore it is reasonable to assume that the growth of cognitive neuroscience and funding thereof is due to a growing (economical) interest in the brain, and a faltering interest in the mind, which at its very crudest can appear to be no more than a arbitarily structured taxonomy of competences.
Sherrington summed up the problem at least 70 years ago:
"Human cognition may like the winged horse take at times its flights toward the stars and forget Earth."
Thanks for your thoughful comment Jonathan. I think that's a useful split you point out between the threads of this discussion.
Overall, though, the discussion in the comments here is quite philosophical. I'm still waiting for some hard evidence that cog-neuro is worth all the attention. For example, if I were to try and write a top ten list of cog-neuro studies as I've done for psychology in general, what would be on there?
Jeremy,
I must confess that I have a hard time understanding the current hostility towards neuroimaging. (Disclaimer: I work with fMRI myself.) Certainly, people have made exaggerated claims about what we can learn about the brain using imaging. But in the end, neuroimaging is just a method - or actually: a number of different methods - for monitoring brain activity. It can be used to learn something about how the brain works, just as single cell recording, neuropsychological testing, genetic knock-out studies, etc., etc. can. I also think it is worth noting that neuroimaging is not only used to locate brain activity. It is also possible to monitor cellular and molecular processes through the use of neuroimaging methods. Indeed, reserachers have recently used MRI to monitor gene expression [1]!
I understand that the question posed by Coltheart is whether neuroiamging can tell us anything about cognitive functions. However, I think Coltheart sets an absolutely impossible standard for answering this question in the affirmative. Why should the criteria be that we are able to decide between (pre-existing) cognitive theories? In my view it is clearly possible to say that neuroimaging has contributed to the understanding of cognitive function, even this criteria is not met.
Having said that, I recently spoke to Peter Hagoort, a prominent psycholinguist, at a conference, and he actually told me that he thinks the results coming out of the last 15 years of EEG, PET, and fMRI language studies rather convincingly shows that Chomsky's specific theory cannot be true! So there...
[1] C. Liu et al. (2007): Imaging cerebral gene transcripts in live animals. J. Neuroscience 27: 713-722.
Martin, thanks very much for your comment, very interesting reading.
I'm interested in what you say about current hostility to neuroimaging. Do you find there's a lot?
On the point you make about Coltheart's impossibly high standards, I think there's some truth to what you say. On the other hand, I think you might be setting the bar too low, by just looking for some (any?) contribution.
Thanks for the language example. There's actually a few companion pieces in that issue of Cortex arguing for what neuroscience has contributed. Perhaps I'll write those up in a future post so we can look more closely at its achievements.