Phunctional Imaging Is The Phrenology Of The Twenty First Century

July 7, 2008

Functional imaging approaches such as fMRI that indirectly detect changes in correlates of neuronal activity in volumetrically defined brain locations are extremely popular right now. Investigators put animals or human subjects in an imaging rig, and then have the subject perform tasks or respond to various stimuli. The idea is that by detecting particular brain regions that become more active during certain phases of a task, one gains insight into how the brain processes information.
This is an open thread for our commenters to discuss whether this set of approaches is a total waste of fucking time and effort that generates pretty picture fodder for making up wackaloon fantasy stories about how the brain works, or, alternatively, is really providing us with information that will lead to a genuinely more satisfying understanding of neural information processing. Go at it!

No Responses Yet to “Phunctional Imaging Is The Phrenology Of The Twenty First Century”

  1. Greg Laden Says:

    Both. My comment will address the former, or at least, the potential for self serving, non-useful proto-wackaloonery.
    The indicator (fMRI pretty pictures) of a phenomenon (brain function details) can only be interpreted if the mechanism is well understood at some level. (There is obviously an interaction between understanding the mechanism and using that understanding to advance the model at hand.)
    It is my suspicion that the current standard model of how the brain works is too tied to the obvious externals of the functionality of the brain than it should be, and what is happening inside the brain is very different.
    Analogy: Two cars, one a suburu forester, the other a toyota prius hybrid. Not knowing how a car works at all, but observing both for a while, one would probably come up with one model to explain the inner workings of both cars. But the inner workings are VERY different. On has a gas and electric engine, with the latter the drive engine much of the time (I think). The other has only a gas engine. One has a transmission that reverses the energy directionality so the car goes backwards (reverse) the other does not reverse using its transmission at all, but rather, runs the motor backwards. And so on.
    The overt functionality of what the human head seems to be able to do is not necessarily mirrored functionally by the way parts are put together inside the mass of neural tissue we call the Brain or the CNS. But much of the interpretation of fMRI assumes this.


  2. It’s totally useless, except when it’s not. Examples: anything involving a human malady that neuroscience has not re-created in an animal model (autism, bipolar, etc); anything involving a human ability that no one has demonstrated in an animal model (abstract language, ability to respond to instructions like “Rearrange these two screen items in your head to decide if they fit together,” etc.)
    I think the only thing we get out of it at this stage is surprising data about brain structures being involved in tasks that we didn’t realize they were involved in. Haven’t there been a few studies showing that there’s a lot of nonmotor language processing in the cerebellum, or something equally weird? (obviously I haven’t paid much attention)


  3. Kevin H Says:

    The analogy of fMRI to phrenology is so tired and ridiculous. First of all, fMRI actually collect real, unbiased data, which phrenology never did. Generally this silly comparison is drawn on something close to the following: Phrenology looks at anatomical bumps on the head, and fMRI draws parrallels between anatomy and activations, therefore they both flawed. However, every current method of studying the brain, from cellular science, single/multi-unit recordings, EEG, and fMRI all accept the basic theory of anatomically derived function, and all have proved again and again that this is by in large a sensible hypothesis. Each of these fields contributes a unique set of benefits and constraints that are useful in getting an overall picture of brain function. The farthest you could conceivably go is to say that fMRI has too much money compared to these other forms of neuroscience, or probably more accurately, that the main stream media has a large bias towards reporting on fMRI because it has pretty pictures.


  4. Kevin H Says:

    Most animal models of Human neurological conditions are flawed. Even if we successfully find the cause of the disease, and there is a parallel in animals, the disease can easily effect a different brain in a different way (for example, what does autism mean without language, and how is this different than autism with language). This doesn’t mean we can’t learn something from them, but this knowledge must constantly be checked against the actual human condition as much as it can, be it through human psychophysics or brain imaging.


  5. bsci Says:

    The idea is that by detecting particular brain regions that become more active during certain phases of a task, one gains insight into how the brain processes information.
    This seems to be your statement about the purpose of fMRI studies, but it also descibes PET, EEG, MEG, optical imaging, local field potentials, spike recording and pretty much any recording in a live animal or even in a dying slice taken from an animal.
    The big differences of fMRI are the exact signal it measures, that you can activity across the entire brain, and that it’s noninvasive.
    If an electrophysiologist could get noninvasive, whole brain data of spikes, I’m sure they would do it. They would probably even make pretty pictures of spiking patterns.
    The real critique of fMRI is the type of signal. The measures of cerebral blood flow or relative blood oxygenation have a larger spatial resolution (2-5mm^3)and a larger sampling rate (1-4s). We know that, even with these limitations, the fMRI signal is tightly correlated to neural metabolism and local field potentials. It’s hard to distinguish events that occur 50ms apart, but we can design many studies that allow for differences can we can see with fMRI.
    Like DJMH says this has allowed us to study entire systems in humans that were previously inaccessible. In many of these cases, the fMRI findings might not provide conclusive results, but they point physiologists to entirely new areas of research they were ignoredt due to the needle-in-haystack approach of electrical recordings.
    The explosion of fMRI has also been paired with a renewed interest in neuro-vascular coupling, an entire area of neuroscience, which has turned out to be much more complex than anyone imaged 20-30 years ago.
    Finally, people like to throw around the word phrenology to discount fMRI research. While there are a non-trivial number of lousy phrenology fMRI papers, we should not forget that phrenology is a real part of neuroscience history that lead to real research in the localization of brain function. fMRI phrenology is showing actual maps of brain activity and a these studies do contribute (not complete) our understanding of how the brain responds to specific tasks.
    In addition, more and more of fMRI are not mapping studies. They are studying networks and how information is stored across voxels. I can write a lot more about this, but a good intro article is “Growing Pains for fMRI” Science june 13, 2008 p1412.
    I have a bit sporadic internet access today, but I’ll respond when I can.


  6. CRM-114 Says:

    Most people think of fear and anger as orthogonal (although that might not be the word they’d chose) rather than opposites. However, they are intimately related in the amygdala, which is vital to both flight and fight.
    If you are the subject being imaged, and you respond to something in the experiment, will the experimenter conclude ‘fear’ when what’s really happened is he’s pissed you off?
    I don’t know of any brain regions that are ‘single-user’. If a dozen different activities can engage the same regions in the same order, then figuring it out is anybody’s guess.


  7. fMRIMHO Says:

    I hate fMRI papers at this current time, but I wonder if technology will actually improve to the point where noninvasive imaging in humans will generate neuronal activity data at a meaningful temporal and spatial resolution, i.e., at the subcellular/synaptic level. Studying live (wake-behaving) humans is the most optimal organism to study behavior. This is probably the only way to study complex things like consciousness…


  8. Ace Says:

    Sigh… I find myself having to defend fMRI as a valuable method in neuroscience more and more often these days. I do understand why so many people are annoyed at this method but it is NOT true that all, even most people in the field are not aware that this is one method, with its strenghts and weaknesses, that it is just but one tool for science and that it should not just be used willy nilly on any topic (“lets do fMRI”) because it’s not well-suited to answering many kinds of scientific questions…
    Full disclaimer: fMRI is one of the methods I use. That means, I find it insufficient on its own to answer all my research questions but I find it a fine method to address some of my questions or parts of my questions. I have also been using it for some time. I wanted to write and put in a comment on how the field has been changing over the last years.
    Yes, there is A LOT of wacked out stuff out there with fMRI. People got excited and did any kind of experiment they could think of and for some time, these got published in high impact journals rather easily due to the novelty factor. Some of these early papers are classics, very intelligent and actually shedding light on Important Problems, some were just fluff and 10-15 years later, in retrospect, it’s easy to separate the wheat from the chaff.
    But believe me, despite the fact that we still encounter some “neural correlates of love handles” or “neural correlates of simultaneous bladder overfilling and rectal stimulation” kind of experiments in journals (and we usually e-mail those citations to each other and engage in geeky commentary/speculations) and despite the fact that there is a crazy amount of media attention to fMRI studies (as well as concern about some of this among scientists, see or sign in to NY Times), the bar for publication of fMRI studies have gone really far up in recent years. You just cannot do one experiment that “lights up” some brain areas and publish it any more, certainly not in any high impact journals. This simply is not true any more. These days, even to publish in non-GlamourMag respectable journals (say, J Neurosci or Cerebral Cortex), it’s customary to carry out multiple, detailed experiments, or data from multiple methods, or you must have a really smart idea that answers a clearly defined Important Question.
    Also worth remembering is that not all of fMRI is used in the service of “cognitive phrenology”. Contrary to popular belief, it’s not just the fMRI papers about aesthetics, empathy and love that get into the GlamourMags. There is a lot of biological research that is continuous with neuroscience research from many decades ago. Issues like mapping of sensory and motor areas for example, have been very active (esp. the exploration of visual areas). And now with fMRI being possible on both humans and monkeys important findings are being made re: comparative neuroanatomy and evolution. Techniques such as multivariate decoding and repetition suppression are often able to help specify functional properties of and computational principles in brain areas. There is also lots of methods and clinical work. As with every method, there is a lot of crap and some small amount of really good work. But overall, the kind of “blobs in the brain” fMRI is not the state of the art in the field so it’s not fair to criticise the field as if this is the case.
    Finally, it is an annoying fact that people are just attracted to brain images and tend to think that if there’s some brain study involved, it must be interesting/true. This helps explain the current media and public attention to the topic. This paper shows that only neuroscience experts (not even graduate students) were immune to the lure of the cool brain images: 🙂
    But isn’t there a similar “awe” towards, say, genetics? To an extent, the hype will pass. And maybe brains with colours and blobs will remain exciting to some degree, but that doesn’t mean ALL scientists are in it for the attention and glory. fMRI is a good method for some kinds of questions and plenty of people are trying to use it (sometimes in combination with other methods) to address longstanding biological questions.
    It is true that there are a lot of misconceptions or misinterpretations of fMRI data. See
    But my overall point is that this is a method just like anything else. It can be used in appropriate ways to the questions at hand, or arguably, in wackaloon, dumb, boring etc ways… Data can be interpreted appropriately or inappropriately. Maybe people got a bit excited about fMRI for a few years, but if anything, there is now a backlash. My view is that it will balance out and we’ll deal with it. Right now the bar is continuously getting higher for fMRI papers and grants. Researchers are either having to pull their weight by doing meticulous, tight experiments addressing Scientific Questions, or they simply can’t publish their work any more…
    Science just is what it is…


  9. bsci Says:

    I mostly agree with you. Many fMRI studies are generating these maps to specific tasks and over time we see that the same area is activated for different tasks and has different roles. This advances our knowledge. The fact that some studies activate the amygdala in response to fear and other to anger tells us it has a role in both.
    A meta-review that takes this issue a step further is “Superior Temporal Sulcus -It’s My Area: Or is it?” Hein & Knight, Journal of Cognitive Neuroscience 20:12 pp1-12. They show five different tasks that have been shown to clearly activate the STS. They how there is some spatial localization within STS, but some of the tasks cannot be spatially sublocalized. Some of the inseparable tasks activate a different pattern of other regions thus it’s more the pattern of activity across a specific set of regions rather than specific region that informs us. Except for perhaps MEG, no other modality besides fMRI could help us build such a model.


  10. DrugMonkey Says:

    fMRI is no different from the list of other gee-whiz techniques in neuroscience from behavioral(ist) rigor, to the initial use of anatomically directed lesions, to some of the first neurochemical tools, to EEG/evoked potentials, to unit-recording, to genetic manipulations, to fMRI, to….the FUTURE!
    At each point of hey-day, the latest and greatest techniques enjoyed a reputation and popularity based on the technique itself that was not justified by the scientific design (in many cases) of the resulting papers. The limitations that should have been obvious at the outset were overlooked, then later became set-in-stone acknowledgments of drawbacks- for some of you young’uns, just think about the history of mouse genetic tools. You should have an appreciation for how we have come forward from “any constitutive knockout and a half a phenotype is good enough for a CNS publication” to the present demands for regulated, tissue (or cell-type) specific techniques with extensive phenotyping, supporting in vitro, etc, etc.
    Each and every gee-whiz scientific bread slicer has gone a similar route, you betcha. When you are getting all excited about how the latest chip-ChIP-CHIPPY-DIPZ-N-SNPz technique is going to TOTALLY REVOLUTINZE TEH SKIENZE!!111!!!! you’d do well to remember your history lessons.
    Yet. Many new gee-whiz techniques do become incredibly valuable members of the neuroscience toolkit, as long as they are used appropriately and viewed as contributing (only) to a specific part of any neuroscience puzzle.
    This brings me around to PhysioProf’s objection to publishing “quotas”. Waaaah. Journals always have quotas for this that or the other kewl technique that tick off those scientists who have more pedestrian stuff of objectively higher quality that can’t get similar high-profile acceptances. So you would have to either attack this as a general principle if you want to be taken seriously. Bitching about one specific gee-whiz exemplar while being all in favor of some other gee-whiz exemplar isn’t that convincing- we can all complain about how our gee-whiz is better than your gee-whiz technique…
    Disclaimer: I bitch about prioritization of gee-whiz over more universal quality indicators all the time, so I’m not trying to holier-than-thou PP’s comment at all. Scientifically I would tend to hew more toward appreciating fMRI over other gee-whiz techniques but I’ve been known to lie down with the enemy as well. As always, it is likely that I have previously held, currently hold or am seeking to hold funding to conduct research including one or more of the gee-whiz techniques under discussion.


  11. neurolover Says:

    Because many of you are reductionists, you’re dismissing the idea of studying things at different levels of analysis. fMRI is useful now, though limited, and much of the coverage of it, especially in the popular press, is ridiculously exaggerated. It’s usefulness does not require it to become the magical method of measuring intracellular activity at a single cell level, simultaneously all over the brain.
    Our answer, to a big question, like “how do I think up and type these words in this comment” is going to come at lots of different levels. One of them might be as general as these areas of the brain seem to be active during the process, as opposed to when I sit still. Arguing that this is meaningless information is no more justified than the physicist arguing that we don’t understand an ion channel because we can’t model the atomic forces.
    And, yes, people get excited about very bad fMRI, but that’s ’cause they care about people, not ion channels, or rat brain slices, or bird brains. The job to do is not to diss fMRI, but to work with naive people so they understand the strengths and failures.


  12. S. Rivlin Says:

    The issue of fMRI in neuroscience is not different from any other technology or technique when first introduced as a possible approach to study the brain. Any new technique goes through growing pains when at the beginning a very small group of scientists uses it. As more publications appear where the technique is the main producer of new data, more scientists are willing to try it and trust it. Sometimes, especially when a technique is responsible for new discoveries, its popularity skyrockets and its users tend to forget its drawbacks. Thus, on occasions these scientists find themselves in a bind as they reach conclusions unjustified by the available data. Many good and useful techniques had become obsulete over the years as new and better ones have emerged. This fact should not discounted the contributions and the importance of old, discarded techniques. fMRI is a technique, which eventually will take that very course. Until then, neuroscientists should make the most of it and be aware of its drawbacks, while making attempts to improve it.


  13. Mo Says:

    The powers of fMRI are grossly exaggerated by the media, but it can be very useful in identifying those at risk of Alzheimer’s, and in identifying structural abnormalities associated with various neurological conditions of childhood.


  14. Ace Says:

    So what’s wrong with my comment guys? I am posting it again separately in case you did not receive it the first time. If you did, why not post it? Is it offensive or something? Is it just too long? I really wanted to have a comment on this as someone who works in that field for many years plus there are some good links in there that readers might find interesting. (Plus I spent 15 minutes typing it up and I hate to waste time… What’s wrong with it, do tell!
    Obviously I don’t expect you to publish this message.


  15. S. Rivlin Says:

    The editors of this blog have found your comment to lack supportive data and are regret to inform you that they decided to reject it. 😉


  16. DrugMonkey Says:

    The editors of this blog have found your comment to lack supportive data and are regret to inform you that they decided to reject it. 😉
    Interestingly, it is the inclusion of supporting data via linked citations that gets comments trapped in the spam filter.
    Readers, as always, if you have a comment with links that doesn’t appear, drop me an email and I’ll fish it out for you.


  17. bsci Says:

    Ace, Scienceblogs has an automatic spam filter if you include more than a certain # of links in a comment (5?). Hopefully the monkeys who run this site will find and publish your link now that they know it’s there. (unless S Rivlin is right)


  18. bsci Says:

    I definitely agree with DrugMonkey and S Rivlin that there is value in fMRI, but it definitely still has a high gee-whiz rating and attracts a lot of new people with poorly designed experiments. Still, specifically for fMRI, it has helped cognitive psychologists, psychiatrists and social scientists think about their research from a neuroscience perspective in a way that was previously impossible or impractical. Merging new knowledge bases into the neuroscience community is changing neuroscience and getting us closer to linking non-human research to human applications/treatment evaluations (i.e. the reason we all get funding).
    That said, the people with no neuroscience/statistics background using fMRI are some of the biggest problems in the field who publish some of the worst papers. Still, more and more people in from these fields are getting training in neuroscience/imaging statistics at earlier stages of their careers and the average quality of papers in these areas is definitely increasing (or I’ve gotten better at skipping bad papers by skimming titles)


  19. Ace Says:

    Dear Rivlin et al,
    Thank you for submitting your rejection of my blog comment “Re: Phunctunal imaging is the phrenology of the twenty first century”. Unfortunately, I receive a high number of excellent rejections and am able to accept a only small percentage of these. I have now received peer reviews from several experts in the field, and while they found your rejection to be interesting, timely, and well-executed in general, they had some concerns regarding its novelty and its interest to a general scientific audience. Consequently at this time, I an unable to accept your rejection. We hope that your rejection will be well-received at a specialist blog and thank you for considering me as an outlet for your rejection.
    Best regards,
    (Hehe, I’ve always wanted to do this dorky rejection of rejection thing..)


  20. Since Mo already plugged his own post on the topic, I will mention several other lengthy discussions of these issues in other neuroblogs:
    Two cautionary notes on the use of fMRI in Small Gray Matters.
    Strengths and Limits of fMRI studies on the brain in Deric Bownds’ MindBlog.
    Science, Criticism, fMRI in The Frontal Cortex.
    The fMRI smackdown cometh in Mind Hacks.
    Logothetis (2008): A must read for proponents of *anything* (and everthing) in Talking Brains.
    Everybody’s a Neurocritic! in The Neurocritic.
    AND a number of commenters here have made substantial contributions to the dialogue as well.


  21. Christopher deCharms talked about this topic on TED. I would say he is a definite advocate of fMRI.


  22. neurolover Says:

    The link between the signal measured in fMRI and brain activity is an important concern (as detailed in the Logothetis review). But, I don’t think that link is the most significant concern about fMRI. What I am more concerned about is sloppy experiment design, as detailed in the controversy over the “political” study, which was apparently published in the NYTimes as its first venue and the statistics (which we haven’t brought up yet). The experiment design requires a strong knowledge of psychological experiment design (which the psychologists have known for a long time). Doing a bad behavioral experiment, and sticking someone in a scanner gets you nothing at all. The statistics require a sophisticated knowledge of statistical analysis, more and more sophisticated as we start crunching larger and larger data sets in more and more complicated ways.
    (incidentally, these second two concerns are also important for genetics & molecular biology — for the phenotyping of genetic defects, as those experiments become more and more complicated, and in the crunching of numbers in genetic analysis and microarray data, and other huge huge datasets).


  23. You know PhysioProf, I think we will continue to need pretty pictures if we ever plan to sell the public on our findings, but I think that most imaging is, as a wise man put it, “a complete fucking waste of time.” So often these measurements are not quantifiable. What does it mean when a part of the brain “lights up” (as I have heard it described on CNN). The answer? Not a mo-fo thing unless you can asign a quantity to it or at least understand the scale of said lighting up. I have worked with people using contrast echo to look at whole organ perfusion and asked them how they quantify changes in blood flow. Their respone? The area gets brighter or darker. I think that non-quantifiable imaging is nice to help us generate testable hypotheses with the potential to the collect quantifiable data, but that it is too easy to hang our hats on a pretty picture.


  24. bayman Says:

    Knowing zero about neuroscience…what about that shit where people lack conscious perception of certain types of stiumli, but still light up the right part of the brain when presented with said stimulus. This kind of result can tell us something, no?


  25. DSKS Says:

    I don’t necessarily think the fMRI hype is such a bad thing. The kind of stuff that plants that initial spark of scientific interest in the public (some of whom constitute the next generation of scientists) is often seen to be fanciful and superficial in hindsight. I agree with the previous comments suggesting that part of engaging the public and recruiting minds necessarily involves pushing the marketable aspects of technology, and purty pictures and movies sure work well at that. I’d sooner the media was fawning over slightly embellished conclusions from fMRI studies than over psychic children or the latest incarnation of Jesus in a pile of horseshit. Besides, even if fMRI studies do not generate particularly strong conclusions, they can be a good source of new and exciting hypotheses, which can be subsequently tested with more rigorous techniques (if not in humans). So I would disagree that such work is a total waste of time.
    Obviously it’s frustrating to see flimsy material getting so much column space, and in an ideal world substance will continue to win over style where it matters, which is within the scientific community and not the techno-tabloid’s armchair readership.
    In an ideal world.


Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s

%d bloggers like this: