Serendip is an independent site partnering with faculty at multiple colleges and universities around the world. Happy exploring!

Imaging and the Question of Consciousness

rdanfort's picture

Neural and Behavioral Sciences Senior Seminar
Bryn Mawr College, Spring 2010
Imaging and the Question of Consciousness

In 2007, the New York Times published some startling news: “based on analysis of brain-imaging data from just a handful of swing voters, they had divined what the rest of the undecided masses truly think about the upcoming US presidential elections.”  John Edwards was subconsciously disgusting to supporters, because viewing his image activated the insula and the insula is associated with feelings of disgust.  If accurate, this finding marked a radical departure from the uncertainty of the poll to dialogue on the hidden and multifaceted workings of individual minds.  The use of fMRI and other functional imaging techniques is often claimed to reveal long-pursued facts about consciousness, from the local origins of love to the emotional content of a sandwich.  Many feel, however, that enthusiasm for finally answering these questions has blinded journalists and researchers to the technical limitations of the technology, commenting extensively on its potential gaps and misdirections.  Still others propose that our theoretical framework is not constructed in a way likely to uncover the causative link between brain and mind, arguing for an adjustment in the goals and methodology of functional brain research.

Background Reading:

Also of Interest:


Some relevant thoughts from last week


Discussion summary (Sarah)

As an introduction to our discussion of imaging, Jeremy and Bobby presented a screen-shot of a research poster entitled “Neural correlates of interspecies perspective taking in the post-mortem Atlantic Salmon: An argument for multiple comparisons correction.” This study claimed that they saw BOLD fMRI activity in a dead salmon after presenting the salmon with a photo. Although this gave us a good laugh, more importantly, it was the jumping off point for a discussion regarding why fMRI data is not completely empirical and the limitations of fMRI. Some of the limitations discussed by the class included:

  1. BOLD technology assumes mapping neurons using the most oxygen is a true map of brain activity (this ignores signal origination and the circuitry of the nervous system)
  2. BOLD cannot distinguish between excitatory and inhibitory events
  3. One only sees activation if there are a large number of neurons using oxygen in the same location, not if one neuron each is using oxygen in disparate parts of the brain. (These are only a few of the many limitations discussed.)

Next, we began to transition to a debate as to whether fMRI is better than phrenology, and whether we can be more or less sure of fMRI data than lesioning or physiology data. Issues brought up by the class included whether fMRI can pinpoint the location for love in the brain (general consensus was that fMRI cannot do so) and the political “mind-reading” of swing voters. The political analysis study claimed that if subjects showed high activation in the insula when shown a photo of a particular candidate, they were disgusted by that particular candidate. Although this study was published in the NYT, it was performed by a commercial outfit that was hoping to sell this “imaging product” to companies as a means to figure out customers’ “true” desires and views. Again, this study was quickly debunked by the class based on the knowledge that fMRI activity does not allow an empirical activity → function interference. Paul, however, then brought up the issue as to whether fMRI/MRI allows for greater confidence than do physiology or lesioning. As a class, we initially claimed that lesioning is more useful, citing patient H.M. as an example. Although patient H.M. allowed the neuroscientific community to   learn a lot about memory, Paul encouraged us to be cautious, and brought up the issue of lesioning the visual cortex. When the visual cortex is lesioned, patients report not being able to see; however, when they are asked to point to where an object is located while pretending they can see, the patients are generally correct. This example compels us to question all of our methods of inquiry, noting both limitations and strengths inherent.

We also delved into the assigned reading, “How to solve the mind-body problem,” by Nicholas Humphrey. The class wondered whether sensation is equivalent to consciousness, or does sensation only exist in the present moment? After some clarification assistance from David and Paul, we concluded that sensation is roughly equivalent to primary consciousness. Humphrey in fact asserted that sensation/primary consciousness necessitates nervous system involvement, and is not simply an input only phenomenon. This allowed us to conclude that sensation is an active process, but that there might not be a demonstrable place in the brain where these sensation/consciousness processes occur.

Lastly, Jeremy asked us whether if we possessed the technology in which imaging could tell us everything about the brain, is this something we would want? Or is this an invasion of privacy? Megan expressed doubt that we would ever reach this point in imaging technology. Others equated this type of imaging to a lie detector (a remarkably more sophisticated/complex, and hopefully, more reliable one at that). Some students believed that this level of technology would be an invasion of privacy, but then others aptly pointed out that we would not sneak up on someone with a huge fMRI machine – they would have to be warned. We continued debating whether or not this technology will actually come to fruition in the future – a study was mentioned where fMRI/MRI was able to draw the shape a person was thinking; however, this was only after the individual’s pattern of thinking regarding particular shapes was learned by the machine/computer. Consequently, these types of technologies would have to be individualized for every person in order to be used to determine if someone was lying, etc. Before the class ended, we briefly talked about the privacy issues, job screening, and laws that would accompany this type of technology- hopefully we will continue this aspect of our discussion next week.

Continuing conversation in forum area below



rdanfort's picture

Interesting point raised

Professor Grobstein brought up something that interested me - "Why do we look at lesion studies as reliable?"  The truth is, as we more or less came to recognize, they are not terribly reliable.  I recall the introduction to Phantoms in the Brain, which pointed out that psychology is very much in its infancy, having only begun to move past the "find a patient with an interesting lesion" stage.  These studies are compelling because the effect is dramatic and observable to just about anyone, but are at a clear disadvantage because of their small sample size, low precision, and invitation to generalize without cause.  In studying a lesioned patient with a neural deficit, there really is little  to tell us whether the entire lesion is responsible or only part of it, or whether the lesion represents a site where deficient functions are normally performed or only part of a messenging system.  These are examples, but there are many other ways in which a patch of brain being destroyed might produce functional disability without being responsible for the function that was lost in a meaningful way.  One may hope that the use of fmri on lesioned patients may focus such studies, but it seems possible that we are reaching a limit to this investigational framework.

I'm not certain where to proceed from here - imaging absolutely has a place in neurobiology, and I can't imagine that we could understand brain function very well without it.  It's mostly a question of how we can connect imaging data to the processes being carried out by the brain.  DTI and other emergent techniques may be an excellent start, but it's good to be cautious given Humphrey's writing.  It may be that even a perfect understanding of the pathways activated by stimuli would not tell us what about the brain produces consciousness or what consciousness means.

meroberts's picture


Imaging seems like a valid tool to study the brain and its activity. Especially because the technology will only get better as science advances. However I agree with the majority of the class that imaging has a long way to go before it can be used to study brains in comparison to other brains. As has already been stated, no two brains are exactly alike. Similarly, as Claire pointed out, a hypothetical "brain-map" of an individual will change throughout their lifetime. For example, an individuals' brain structure will change as a result of a traumatic head injury. As it is now, imaging techniques are unable to definitively prove that any one region of the brain is solely responsible for a particular cognitive process. There is no such thing as a standard brain to use as a baseline, unless the size and structure of many brains are compared in order to find an average.

mrobbins's picture

Can the truth be black and white?


          I agree that fMRI has a long way to go before its findings can be
completely understood. We just have to wait for the technology to catch up
with us. However, the world isn’t exactly waiting patiently. As we’ve seen
they are already trying to link its findings to certain emotions,
consciousness, and using it as a lie detector. Two companies have even
gone commercial using BOLD fMRIs to be used as lie-detectors. What is most
interesting to me about fMRI is not what it is necessarily good or bad at
doing, but how its ambiguity is taking the media and even taken the legal
system by storm. The uses of fMRI have become darling to the media.
Currently, a number of legal cases are trying to use fMRI readings to hold
up in court. For instance, there is an ongoing sexual abuse case in which
the defendant is trying to prove their innocence by an fMRI lie detector
provided by a commercial company, No Lie MRI. This company claims that
they can prove the defendant’s innocence. Controlled scientific studies
claim that fMRI lie detectors have between 76-90% accuracy. However, these
studies did not take place in real world settings with different
pressures. Nevertheless, it will be very interesting to see how this case
holds up in court. Even if fMRIs become the next big thing in lie
detection, they are nowhere near foolproof as scientists say they are
still easy to trick. You just think of random things, moving a little, or
holding your breath and the results are rendered worthless. But I guess
the jury’s still out on whether or not the strengths and weaknesses of
fMRI lie detection still beat the status quo. Personally, I’m holding onto
my vote a little longer.


Claire Ceriani's picture

Lie Detectors

The idea of using imaging as a lie detector is intriguing, but I find it very unlikely.  I suppose if we were ever able to truly map the activity of every neuron and know what each one is responsible for, we could use imaging to read minds, but I don't see how such technology is possible.  We know that brains are capable of changing dramatically in reponse to certain situations.  Even if we were able to map a person's brain, I don't think we could assume that the map would continue to be valid for very long.  Individual neurons may be involved in so many different processes that their function may shift and change very quickly as new experiences are processed and new information is learned.  While we can determine which areas of the brain appear to be involved in certain processes, I think using imaging to understand individual neurons will never be a practical reality.

LMcCormick's picture


  I am intrigued by the concept that memories may exist as small changes distributed throughout the nervous system.  If this is true, then the possibility of using fMRI as a lie detector seems unlikely.  While it may be possible to use fMRI to predict what a person is currently looking at (as with the study we mentioned in class), this is a much more basic task than identifying memories.  We may have defined (individualized) “maps” of activation when looking at a picture of a house, for example.  However, recalling a memory may induce very different brain activity than the activity created while receiving real-time visual input of an image.  Because imaging techniques look for areas of strong localized activity, it seems unlikely that fMRI would ever be able to detect a subtle and diffuse signal characteristic of recalling a particular memory.  Because the practical use of lie detectors is to question someone’s memory (not whether they are currently looking at a picture of a house), it seems improbable that fMRI would be very useful in this context.

Bo-Rin Kim's picture

It is difficult to conclude

It is difficult to conclude what we can draw from brain imaging studies, as we are not sure what we are actually seeing. Are the lit up areas involved in more activity or are we simply detecting the accumulation of molecules that do not speak at all to the amount of cognitive activity that is taking place? Moreover, as Professor Grobstein mentioned, there is the problem of individual uniqueness and the inability to generalize brain-imaging studies to the general population. However, as Alie and EB mentioned in their posts, I think that fMRI, EEG and other brain-imaging techniques, while not perfect, are a good place to start learning about how the brain functions. While fMRI may not give us an accurate portrayal of what brain regions are involved in certain behaviors, it gives us a starting point to investigate further. Moreover, when techniques like fMRI are cross-referenced in their findings with other techniques like EEG (showing temporal brain activity that matches up the spatial findings of fMRI), it adds credibility to the brain-imaging findings. While correlations in imaging to behaviors cannot conclusively determine that certain brain regions are responsible for certain behaviors, it helps us to gain a better general sense of how behaviors and the brain are related. In addition, the reproducibility of brain-imaging results across different people shows that these images are showing something significant. They may not prove that these brain areas are responsible for/the origin of certain behaviors, but they show that are at least some how involved, and their involvement is similar across people.

Paul Grobstein's picture

brain imaging, brain research, and humanness

Lots of food for thought, both in last week's session and in the follow up discussions here.  Some things that stick in my mind for continued consideration ...

I share a sense that imaging studies, like all studies, have both uses and limitations, and like the idea that what we can have most confidence in is what is supported by observations using multiple techniques (cf EB, Sarah).  At the same time, the discussion highlighted some deeper problems inherent not only in imaging but also most other studies with which imaging studies might be compared.

One aspect of these problems has to do with a localization presumption, and an associated inclination to average and to attribute special significance to locations that display relatively high average activity.  There is no assurance whatsoever that the brain is actually organized in ways that fit these presumptions.  It is entirely possible, indeed even likely, that diffuse activity, of a kind that will not be detected as localized high average activity, plays a significant role in many nervous system processes.  That we can detect areas of high average activity correlated with particular behaviors provides no evidence whatsoever that those are the critical substrates for generating those behaviors. 

The history of the search for "memory traces" provides an apt case in point.  Reviewing his extensive program of lesion studies, Karl Lashley wrote in 1950 in his classic "In Search of the Engram"

"It is not possible to demonstrate the isolated localization of a memory trace anywhere within the nervous system"

The situation has not changed since, despite extensive and ongoing research using a variety of techniques.  Indeed, the upshot of subsequent research has been to establish fairly convincingly that in general "memories" involve  distributed small changes that are widely distributed throughout the nervous system and that they are, to a large extent, constructed and reconstructed in real time.  In short, there is no "engram" of the sort that might be detected as localized high average activity.  Areas exhibiting high average activity in imaging studies, like those that "abolish" memory in lesioning studies or those that yield memories when stimulated electrophysiologically , are more likely to be involved in constructing memories than to be the storage depot for them. 

The point here is not that "localization" doesn't exist in the nervous system.  It certainly does, but what is "localized" rarely bears any strong relation to the functions we are inclined to believe exist from observations of behavior.   There is lots to learn from imaging studies (Allison), as from lesion and electrophysiological and pharmacological ones.  What needs to be kept in mind in all cases is the importance of doubting findings that fit neatly with what we presume we ought to be looking for, and so have developed the techniques to find. 

The other deeper and more general problem that our discussion of imaging technologies seemed to me to highlight is the presumption that particular patterns of brain activity should generalize across people, and across particular times and contexts for individual people.  Why should we think that either "lying" or "redness" corresponds to the same pattern of activity in different people, or even to the same pattern of activity in the same person at different times?  They might, of course, and its certainly worth asking the question given available techniques, imaging included (cf David F).  But its the sort of question that should be asked skeptically, with full awareness of the limitations of techniques and the bias to find what one is looking for.  What one should be looking for is not the expected but the unexpected. 

Along these lines, my own sense of the degree to which the brain is continually constructing and reconstructing makes it entirely possible that "lying," "redness," and most other things of this sort are actually a lot like the river Mark Twain describes in Life on the Mississippi.  What makes the Mississippi the Mississippi is not that it has the same pattern at any given time; it is constantly changing with the time of day, with the seasons, and with the years.  The notion of the Mississippi is a construction of the brain, an effort of the brain to create some constancy to make sense of ongoing change.  Given that, my own guess is that the Mississippi as well as lying and redness are related but not quite the same thing in different brains and related but not quite the same thing in a given brain at different times (Vadilson, Vidya)

We can certainly try to "calibrate the fMRI to any individual" and look for "generalization across people" and learn something about the brain from efforts to do so.  Perhaps though what we'll learn though is that there always remains an uncalibrateable and non-generalizable individual distinctiveness, and that this is part of the essence of brain function?  If "fMRI results differed from the subject's verbal accounts" (David F) then the fMRI results are incomplete, reflecting activity in some part of the brain other than those involved in verbal accounts, and an associated internal disagreement   

If so, we can stop worrying about threats of  "mind reading."  The problem isn't that others can see what we are in all its complexity but instead that people might assert that we are things different from and simpler than what we actually are.  That's a problem that long predates imaging.  Maybe its a problem that good brain research, including imaging, might actually help to solve?  

kenglander's picture

While I want to agree that

While I want to agree that there is an "uncalibrateable and non-generalizable individual distinctiveness" that exists in each person, I am a bit confused about how scientists can arrive at that conclusion when they are using techniques that are supposed to be generalizable across subjects.
When researchers collect data using fMRI and other neuroimaging processes, they inevitably come across individual differences. However, these differences are often overlooked (and potentially removed from later analyses) in order to see general trends. When differences are mentioned, they are usually generalized such that two groups-- defined by the researchers-- are compared and contrasted; the individual is still not recognized for his or her unique brain activity. Even if the researchers did discuss fMRI data unique to an individual patient, what would that accomplish?
In a similar vein, if we find individual differences between brains, what do we do with that information? Is there anything we can do with that information? Do we keep trying to calibrate, calculate, and define these differences and why they exist, or do we simply let it go? It seems as though the former is more in line with the general pursuit of science, but if we are defining our distinctiveness as non-generalizable, then is that quest futile?
Perhaps we need to understand the activity common to most brains before we start delving into what is individual?

EB Ver Hoeve's picture

  I agree with Sara in that


I agree with Sara in that most imaging studies are probably somewhere in the middle between the extremes of absolute scientific accuracy and complete fiction. And yet, I felt like last week’s class was mainly spent as a bashing session on fMRIs.  I honestly wonder about the usefulness of that mind frame. As stated, already, in the forum, neuroscience tools have helped us gain a much greater understanding of the brain. No technique is perfect. And as scientists/researchers, I think we would almost all agree that the best data is data that can be supported by different techniques. Do we get the same type of result when we use an antagonist as we do when we ablate?  If yes, then we feel more confident that our findings hold true scientific value. Scientific advancement takes time. Is it really useful to spend the majority of our time critiquing a technique because its accuracy isn’t yet up to our… standards?  Our expectations?  Who are WE? I agree that it would be a mistake if the scientific community became obsessed with fMRIs to the point where they believed the technique to be completely fool-proof, but who is actually doing that? Who, from the scientific community, is actually coming out and saying that fMRIs are perfect?  I don’t think anyone is. Brain imaging techniques have taught us about brain activity. Imaging studies can be used, not as the only way, but as another way to understand the puzzle.


sberman's picture

Imaging study examination

I think that when we look at fMRI and MRI results, we need to pay close attention to what the authors are claiming to show, the method of collecting the data/recruiting subjects, the method of interpreting the data, and the motives of the authors. Although these are factors that must be taken into account for all scientific studies, I believe that this is critically important in imaging studies - more so than in standard research studies, in which  there is more of an established method of interpretation. For example, although the study of swing voters is undoubtedly flawed, there are many clear warning signs to alert us to this fact. For instance, the study was performed by a company looking to market a product, and sought to assign an emotion (disgust) to a particular, highly defined brain region (the insula). However, last year, the results of a study that found that that descendants of depressed individuals had significant thinning of the cortex was published in the NYT. ( Although this study uses imaging data to make its conclusions, it has striking differences from the swing voters study. The study presented in the NYT did not try to claim that the cortex was "the location" for depression - rather, the researchers noted the correlational differences between descendants of depressed individuals and descendants of non-depressed individuals. The authors do not attempt to say that this thinning is an empirical cause of depression. They discuss many possible interpretations of the research, such as that the thinning could be a genetic marker for depression, or could be the result of growing up in a household with depressed parents. Furthermore, the study has a large sample size and is the compilation of generational analyses and data collections completed over the course of 27 years. I feel that these two studies illustrate that there is a wide gap in the quality and utility of imaging studies, and we should examine each individually before saying that it is either an amazing contribution to science or a complete flub - most imaging studies are probably in the middle.

aliss's picture

In class we spent a lot of

In class we spent a lot of time discussing the downsides of fMRI and imaging technologies.  However, I think it is important to remember that these are important tools that have helped us to gain a much greater understanding of the brain.  As long as we fully understand our results as what they are and not as a distinct map of all brain activity related to a given stimulus, or as a way for us to "read minds," fMRI is an extremely useful tool.  Despite the fact that MRI shows voxels that are much larger than a single cell, it still provides good spatial resolution for activity in the brain.  EEG, which has an almost immediate temporal resolution, lacks in spatial resolution, but it still has provided insights into the different roles of different brain areas.  Despite the generality of all of these imaging methods, they have still taught us about brain activity.  While it is important to remember not to veer towards "neo-phrenology," it is also important not to discount the findings of imaging technology studies.

As far as using fMRI as a lie detector, I don't feel that it is less ethical than the use of a regular polygraph.  While that may not be ethical itself (it's often wrong), if we could get an MRI to work as a good lie detector I could support its use in courts.  Since that seems to be one of the only uses of lie detectors in general (ie, there aren't lots of home lie detector systems on the market - or are there?), I think it's within the bounds of what is ethical.  Of course, I wouldn't support the use of MRI as a lie detector by a totalitarian government or anything of the sort, but I don't think the technology is headed in that direction.  While there are plenty of articles that claim that MRI is being used for "mind reading," I believe that these articles are really trying to make the story more sensational.  What these scientists seem to be doing is finding different patterns of brain activity that occur in different situations, and studying them to see how the brain processes different kinds of information, not reading minds.

David F's picture

fMRI as a mind-reader and a theory of consciousness

 One of the interesting discussion directions was the consideration of the ethics of fMRI. If we could develop the kind of technology to read people's thoughts, would we want to? What would we do if the fMRI results differed from the subject's verbal accounts? Which would we believe? Which should we believe? One response that the class offered was that we may never have to face this problem; perhaps fMRI technology will never advance to the point that forces us to make these difficult decisions. As I vaguely alluded to in class, however, this limit may not be obviously true. One study, entitled "Visual Image Reconstruction from Human Brain Activity using a Combination of Multiscale Local Image Decoders" was published in Neuron in 2008 (this may partially address Vidya's concern about fMRI results being taken seriously amongst prestigious journals), and can be found here:

This study illustrates the fMRI's capacity to reconstruct visual stimuli based only on brain activity. Another, entitled "Predicting Human Brain Activity Associated with the Meanings of Nouns" was published in Science in 2008 and demonstrates that this technology can be extended from simple images to object nouns:

However, as Professor Grobstein pointed out, these findings only apply to a single individual; perhaps this capacity can never be generalized across multiple individuals. Although I mentioned that I thought other studies had begun to accomplish this generalization, I had only misremembered another study. So in fact, fMRI technology, for now, is limited to "mind-reading" single individuals. But the force of these studies seems to remain: we are coming surprisingly close, I think, to facing the dilemmas associated with mind-reading. What is stopping us from applying these findings to the public, such as in law settings? Can't we just calibrate the fMRI to any individual in the same way that the studies did to their subjects? And even the generalization across people does not seem, in principle, altogether impossible: Although there will always be individual differences, don't we already know that groups of people can demonstrate similar activational patterns in response to a given stimuli? At the very least, I think we should take seriously the ethics of using fMRI to investigate one's consciousness.

I will also be interested to hear people's thoughts on the Humphrey article. He proposes that our ability to perceive qualia (e.g., the "redness" of an apple) consists in a sort of feedback loop that has become increasingly centralized in the brain over evolutionary history. Thus, rather than having to respond to our bodily motions to infer our experience, that "experience" can exist only in the head. As neuroscience students, I think we are in a unique position to explore whether that is a plausible explanation. Could a resonating loop of neural circuitry (he sets us up to find this explanation intuitive by comparing the cognizing of qualia to the holding pedal on a piano) accurately reflect the perception of one particular aspect (e.g., redness) of a given object?


VGopinath's picture

fMRI and "Pop" Psychology

      One aspect of the issues with fMRI I think that Vadilson was alluding to is how easy it is to overgeneralize maps of the brain.  I have been thinking about how the nature of cancer therapy is going to radically change to be completely personalized as scientists learn more about the vast differences among cancers and the genes mutated in specific cases.  Similarly, I think that once we realize that we can't juxtapose two different people's brains to determine if they feel similarly or differently about an image, the use of fMRI may be used for "pop" psychology.  All of the exciting uses we have been thinking of for fMRI- like lie detectors, focused advertising, determining personal feelings about political candidates- may not be as cost-effective or simply effective as we believed, leading major scientists to look elsewhere.  Hopefully, as the flaws in fMRI become more well-known, colorful pictures of brains won't be sufficient to seduce editors of major newspapers and magazines into publishing semi-scientific results although, honestly I think it may be that they are the only ones who will publish that data. It's unfortunate that as scientific technology has progressed and become increasingly specialized, there's a separation between the data and results that are presented to the public and those that are published in scientific journals.  Perhaps fMRI will become relegated to newspapers and magazines but not accepted as convincing enough data for Nature or Science publication.  

vpina's picture


 Looking at fMRI to determine whether someone is truly going to vote for someone or not seems to be a waste of time and money since we still have not pinpointed a general picture of what parts of the brain mean exactly what. We might have figured out that a general area fires up for emotions but without studying the individual on a personal level there is no known specific location for anger in particular verses joy. Also I’m wondering on a more practical level what this would achieve in the long run?