The Frontal Cortex

Where Are the Surprises?

I was talking to a neuroscientist the other day and he started complaining about fMRI studies. They are too easy, unreliable, etc. (This is a surprisingly common complaint among neuroscientists who rely on the techniques of molecular biology.) But then he asked me a question that I couldn’t answer. “Tell me one brain imaging study,” he said, “that was really, truly surprising? You get lots of studies showing that the DLPFC is important for deliberate reasoning or that the amygdala responds to negative stimuli, but is that surprising?” He went on to note that, while fMRI studies have certainly fleshed out our understanding of cortical anatomy, they hadn’t really overturned any big ideas. Where, he wondered, are the Hubel and Wiesels of brain imaging?

Can you think of any brain imaging experiments that generated some really surprising results? I’m sure there are some, I just couldn’t come up with any of the top of my head.


  1. #1 mikem
    January 16, 2008

    How about the Cambridge study on the vegetative patient which revealed brain activity? That seemed to be rather shocking. But perhaps this is not the kind of study you had in mind…

  2. #2 Alex
    January 16, 2008

    Personally, I think the brain imaging studies of cross-modal plasticity (by Sadato, Pascual-Leone, etc.) showing that visual areas are activated by tactile and auditory input in the blind are pretty interesting and surprising.

    Also, the Kamitani & Tong / Rees & Friston decoding studies are unassailably cool, but maybe not that unexpected.

  3. #3 Jonah
    January 16, 2008

    Thanks. Those are both excellent suggestions.

  4. #4 Stanley Braganza
    January 16, 2008

    Hello Jonah!

    Love the blog, keep it up!

    I had the pleasure of doing my undergraduate degree in Psychology at University College London, and worked under Prof. Patrick Haggard.

    He was hunting for the source of “volition” – where in our brain is free will? (in contrast to forced will – i.e. performing an action because you are told to do so)

    I haven’t kept in touch, but from what I remember, I think he mentioned a study that apparently managed to isolate it…I don’t know if this counts as surprising…

  5. #5 Steve
    January 16, 2008

    This is a shameful opinion for a scientist who clearly hasn’t spent much time reading the literature.

    There are dozens if not hundreds of papers doing a lot more than anatomical localization.

    For example… in the study of consciousness (only because I have a list of papers next to me right now):

    Morris, et. al. (2001) Differential extrageniculostraite and amygdala response to presendation of emotional faces in a corticall blind field. Brain 124, 1241-1252

    Tong, F. (2003). Primary visual cortex and visual awareness. Nature Reviews Neuroscience, 4, 219-229. (this is a review but obviously talks about a number of good studies using fmri)

    Ro, et. al. (2003). Feedback contributions to visual awareness in human occipital cortex. Current Biology, 13, 1038-1041.

    There are a lot more just in this subfield.

    Is there a reasonable way of studying the neural effects of many human psychological properties without fMRI? You can’t do it in monkeys, even intracranial recording in humans is limited in scope.

  6. #6 Jonah
    January 16, 2008

    I think Steve makes an important point. fMRI sure isn’t perfect, but it’s the best window we’ve got. by far.

    I’m sure the scientist I talked to would agree with that sentiment. But he was trying to point out that the nature of fMRI research might not be best suited for surprising research. Of course, the Hubel and Wiesel standard is an impossible standard for any scientist to meet. (For one thing, there haven’t been that many discoveries in any genre of neuroscience that can really compete with their seminal work in the visual cortex.)

  7. #7 JC
    January 16, 2008

    Thanks for posing the issue!

    This warrants a bit of cognitive analysis for what would make a study ‘surprising’ in our modern state of knowledge as a scientific community. (I’ll leave that to you Jonah.)

    Relatedly, it would be a shame to take for granted how far the understanding of the brain has come, to lose the fascination of the type of content that can be accessed with neuroimaging. One avenue of interesting research has to do with reading disabilities and interventions: do brains of dyslexics “normalize” post-treatment or become more proficient at bypassing alternative pathways? Also, studies on blind or deaf individuals whose cortex dedicated typically to the senses they don’t have access to adapts, and becomes used for the senses they do have (a rough summary of the work). How is that for surprising…
    There is so much to still wonder about!

  8. #8 SB
    January 16, 2008

    One surprising neuroimaging finding is that the activity of neurons in visual area V1 is increased when subjects are attending to the location the neurons encode. 10 years ago, single unit recording and ERP studies had failed to find V1 activity and it was assumed that there was none. Now, because of the neuroimaging results, changes in V1 activity due to attention is no longer controversial.

  9. #9 lanoxa
    January 16, 2008

    SB: I am surprised by your statement. Do you know people who are convinced by the fMRI studies? I don’t think neuroimaging studies can resolve that controversy.

    Yes, a number of studies have shown that BOLD signals in V1 are modulated by attention. However, a number of neuroscientists believe that fMRI measures synaptic activity, not spikes. So fMRI can’t tell you what V1 neurons are doing – only what their inputs are doing.

    For example, a classic study by Logothetis found that the BOLD signal is sustained for the full duration of a visual stimulus, whereas the neuronal spike response is very transient (their Figure 3). More recently, Viswanathan and Freeman found that a visual stimulus can increase the tissue oxygen concentration in V1 even when there is no change in neuronal spike rate. (they used a stimulus that causes the LGN to produce very high-frequency inputs to V1, which V1 neurons then filter out).

    So, just as V1 neurons filter out high frequency visual input, they might also filter out top-down attentional signals. Only single-neuron techniques can tell for sure.

    Neurophysiological investigation of the basis of the fMRI signal
    Nikos K. Logothetis, Jon Pauls, Mark Augath, Torsten Trinath and Axel Oeltermann
    Nature 412, 150-157 (12 July 2001)

    Neurometabolic coupling in cerebral cortex reflects synaptic more than spiking activity
    Ahalya Viswanathan & Ralph D Freeman
    Nature Neuroscience 10, 1308 – 1312 (2007)

  10. #10 bsci
    January 16, 2008


    I think you indirectly get at one of the most surprising findings from fMRI. Yes fMRI probably measures synaptic activity rather than spikes, but this measure has discovered a lot of interesting patterns of neural activity that were never observed with spikes. In some ways, it has forced a reassessment of the centrality of spike measurements. The past 10 years has seen an explosion of studies using local field potentials and invasive optical imaging partially because the neurophysiologists are starting to realize how little is learned just by studying spikes.

    Other parallel research that uses and is inspired by fMRI is the better understanding of the amazing process of neurovascular coupling. There is a rich literature both using fMRI and responding to fMRI studies that has radically alter the way we understand cerebral blood flow.

    Dozens of fMRI studies have helped us learn things about the human brain that we hadn’t previously known. I’m not sure what the original questioner would consider worthy of “surprise.” Still, how many years did EEG and then electrophysiology exist before people like Mountcastle, Hubel and Wiesel pieced apart the key elements of the visual system? The concept behind fMRI was discovered in ~1992 and the first serious neuroscience studies started appearing around 1997. Not to mention, we now have so much more knowledge that the bar for new discoveries is higher.

    For that matter, when was the last time someone discovered something truly surprising using only spike recordings?

  11. #11 Geoff
    January 16, 2008

    “Is there a reasonable way of studying the neural effects of many human psychological properties without fMRI?”

    Sure–fMRI is in no way the sum total, or even the gold standard of cognitive neuroscience methods.

    ERPs (event related brain potentials), which are time-locked EEGs. They have very poor spatial resolution, but fMRI has very poor temporal resolution, so ERPs can answer a whole set of questions that fMRI can’t touch.

    Then there’s TMS (transcranial magnetic stimulation) which uses magnetic pulses to temporarily stimulate or turn off areas of the brain. It works great on the motor cortex, and it is being used for other purposes too.

    Going classical, brain damaged patients have been used for years to examine selective deficits.

    And then there are newer techniques, like optical imaging and MEG, which I think provide good temporal and spatial modeling.

  12. #12 bsci
    January 17, 2008

    Geoff, Just to be slightly more accurate, TMS does not turn off areas of the brain. It adds noise. This often disrupts a regions normal brain function, but it’s important to distinguish this from turning an area off since there is are still spikes entering and leaving the region.

    Optical imaging only works on the cortical surface nearest the skull and it is still very much in development. MEG localization is also best closest to the skull and the deeper in the brain the weaker the signal. Probably the real gold standard is psychophysics combine with brain lesion studies which has been around for 150 years, but has significant limitations on both who can be used and what can be interpreted.

    fMRI fills a gap of non-invasive imaging of the entire brain with sub-centimeter resolution in healthy humans that no other tools fills. The best studies are paralleled with other methods, but fMRI is still key.

    To parallel the Hubel and Wiesel question from above, I’d say we are now more in the Hodgkin-Huxley period where we’re getting a better idea about what we’re measuring with this tool and what it means. Hubel and Wiesel’s work was about 10 years later.

  13. #13 Jonah
    January 17, 2008

    I just wanted to thank everyone for their informed and intelligent comments. This is a great discussion. I especially like bsci’s point about fmri still being in the Hodgkin-Huxley phase…

  14. #14 SB
    January 17, 2008

    For anyone intereted in the example I cited above – surprising fMRI results in V1 during attention, there’s a review by Kanwisher and Wokciulik in Nature Reviews Neuroscience (2000) that is a good first stop.

    Also, regarding the Logothetis paper, I don’t think their conclusion was that fmri measures only synaptic activity. I haven’t looked at the paper recently, but I think they reported that fmri response correlated nearly as well with MUA (~spikes) as with LFP (~synaptic activity).

  15. #15 Ianoxa
    January 17, 2008

    SB: Oops, upon re-reading the Logothetis paper, I now see that they only dissociated LFP/BOLD from MUA at a small number of recording sites. That is, LFP/BOLD always had a sustained response, but for 1/4 of sites, MUA had only a transient response. I guess that is not so convincing; perhaps at those sites the electrodes simply missed the local population of ‘sustained’ neurons.

    As for surprising things, I am surprised that fMRI works at all. Why does the vascular system have to be so responsive to neuronal activity, so amazingly fast and precise? It isn’t obvious to me…why does the benefit of fast and precise blood flow outweigh the energetic cost of coordinating it?

    As a case in point, I was once told that the vascular system responds to neural activity by sending an enormous excess of oxygen, far more than is needed to nourish the tissue. This is advantageous for reasons of speed: a high concentration of oxygen in the blood creates a steep concentration gradient so that oxygen can enter the tissue more quickly. In other words, the brain is willing to exchange a lot of shipping effort for a little bit of speed. Neuroscientists should be thankful for this, because otherwise the BOLD signal would be very weak, and fMRI would be impractical.

    Can anyone confirm whether this story is true? If so, why does the brain have this need for speed?

  16. #16 SB
    January 18, 2008

    One explanation is that when you increase blood flow, you also increase the speed at which the blood moves through the capillaries. The increased speed means that the odds of a hemoglobin molecule giving up its oxygen to the capillary bed go down. So, the vascular system over-supplies the tissue to compensate for the reduced extraction rate. But I don’t follow this literature, so I don’t know what the current thinking is.

  17. #17 bsci
    January 23, 2008

    I just noticed Ianoxa’s question. If you’re still looking here and interested, I can write a more details reply. In short the descriptions are accurate, but the key is that vasodilation isn’t controlled by oxygen concentration. Multiple factors including glutamate update, Calcium concentrations and blood CO2 levels (blood pH) are all more directly related to whether or not blood flow increases. The O2 increase that we observe with BOLD does bring in more O2 to the tissue is mostly a side-effect of these other factors.
    Research in this area of neuro-vascular coupling is still very active.

  18. #18 nick
    January 24, 2008

    Well according to scientists mapping the brain over at Harvard new and exciting discoveries in nueroscience are just around the corner. Check out the link, this is going to rock the world once the new-neurocartographers are finished.

New comments have been disabled.