Saturday, October 23, 2010

Introspection... happens in the brain!

Neuroscientists find some anatomical correlates of accurate self-knowledge.

One of the few remaining bastions of theism is the mystery of the mind. That, and the fundamental nature & origin of the universe, comprise the only serious mysteries from which theistic world views take more-or-less justified comfort. With respect to the latter, a recent scientific American article (& book) by Stephen Hawking and Leonard Mlodinow takes a decidedly non-theistic view, while also claiming that there may be no final theory of everything, only a set of partially compatible theories. This would be like settling for the conflict between quantum mechanics and relativity+gravitation and saying "we're done now". I'm not sure how much sense that makes, since reality is unquestionably unitary and happily doing its intrinsic computations all the time at all scales. So they have to be compatible, and it is only our conceptual apparatus that has failed (so far) to devise a model that is as unified as that reality.

The ultimate problem is likely to be our complete inability to devise empirical tests for such ultimate models of reality, since we can't attain the energies of the early big bang, let alone the putatively timeless, space-less conditions that might have preceded it. Thus we might ironically find ourselves back in an intellectual world reminiscent of theology, another scholarly pursuit which imagines what could be without the benefit of experimental backing.

Sorry to digress, but these topics are closely tied for those unwilling to find meaning in this mundane disenchanted world, and among their fellow bio-people. And for whom mystery (i.e. lack of understanding) is the essential ingredient for meaning. Which is to say, provides the space into which can be poured projected meanings.

But back to the mind. That personal, magical place that makes the world appear, and throws in dreams for free. To scientists, it is a problem, but for theists, it is the rock of mystery, the self-evident communication portal to the "other side", and source of untapped magical powers (dianetics, prayer, etc.; see my soul posts.)

Unfortunately, as far as actually known, the mind is confined to the brain- which is exceedingly complicated, but also finite and accessible. Thus it can and is being progressively disenchanted by scientists doing their mundane thing. The current paper is one more tiny step on the long road to figuring out how the mind works, now looking at brain correlates of our ability to know how well we actually know what we think we know.

Introspection is one of those prized aspects of human-ness, developed into a high art/skill by the Buddhists, and generally recognized as a key to a better life, whether self-examined in the mode of Socrates, self-analyzed in the mode of Freud, or self-criticized in the mode of Mao(!). We assume that animals are less introspective. Language may be critical to this level of thought, but of course it is hard to tell.

It seems essential to have a way to gauge one's accuracy in all sorts of actions and thoughts, and on top of that to have that way be conscious, not just a form of unconscious training like learning tennis through physical practice. This paper looks at the question in the simplest terms, measuring accuracy of a simple visual task where people vary both in ability and in the meta-ability to know how good they are, to a significant degree. In part it was motivated by brain damage studies showing that a region called the prefrontal parietal network is necessary for this kind of introspection/metacognition.

"We hypothesized that individual difference in metacognitive ability would be reflected in the anatomy of brain regions responsible for this function, in line with similar associations between brain anatomy and performance in other cognitive domains."

Visual test by dim patches, one of which was made darker than the others. Test subjects were then asked their confidence in calling which set (one or two) and which patch (one to six) was that darker one.
The visual task was to compare two sequentially presented sets of "gabor" patches (Fig 1) and say whether one of the patches in one of the sets was darker than the other  patches. The task could easily be tuned in difficulty to make every subject perform at 70% accuracy. Then they asked the subjects to rate their confidence about each choice, from one to six. The figure below shows the ROC curve for each of the 36 participants, relating false positive rate (confidence level when wrong; horizontal) to the true positive rate (confidence level when right; vertical).

Individual ROC curves for all participants, showing how well (area under the curve) they introspected their own performance in visual identification.
As is customary with ROC curves, the curve bows up as the process is able to pick out signal from noise at a better than random rate. In this case, if all subjects got their confidence completely right at all times, the lines would shoot from zero to one at the origin and stay at one the whole way going right. That would lead to maximal area under the curve, (over the diagonal, which is random performance), which is the typical measure of signal-to-noise accuracy in many fields.

Ordered display of the graphs above, relating actual performance (blue) and introspected performance (red).
The researchers found extensive individual variation among their 36 test subjects, both in performance ability and metaperformance- the accuracy of their confidence levels. The graph below gives the latter data, displayed in order. The real question was then... can we correlate this variation with any anatomical aspect of the brain? They used MRI to look all over the brain at the volumes of gray matter, which are the neuronal cell bodies. And they also looked at the white matter connectivity, which uses a new form of MRI that allows users to trace myelinated pathways in the brain. This has been a powerful addition to the functional and anatomical MRI toolchest.

Example of the data available through white matter imaging, using diffusion tensor imaging, or brain "tractography".
Looking first at the gray matter, they see a few spots of correlating variation (red in the pictures below, mapped on inflated brains). It is quite remarkable that such a method could point so specifically to particular areas of the brain, since they just cranked the entire brains of all participants through a complex normalization, inflation, and statistical area comparison method. One would imagine that a more complex picture would emerge, especially in view of the modest nature of the underlying correlation, shown in B. The Aroc graph in B shows the real signal- correlation between test performance and portions of Brodman area 10. For comparison, the right graph (d') shows the correlation of the same areas not with the metacognition performance, but with actual visual performance, which serves as a negative control- no correlation.

Volumetric data, showing islands of grey matter mass correlation (red) and anti-correlation (blue) with introspective ability. Graphs show island correlation (Aroc) and control of actual visual ability (d').
Needless to say, this section of the Brodman area matches that previously suspected to play a role in this form of meta-cognition from other studies, including functional MRI, disturbance with transcranial magnetic fields, and brain lesions.

Lastly, they also probed white matter pathways, (with MRI "tractography"), and found what looks like an slightly more significant correlation between metacognition and an area in the corpus collosum which allows communication between the hemispheres. The identified area is specifically linked via wider brain wiring to the Brodman's area mentioned above. Unfortunately, the graph fails to offer error bars, so the significance of all this is somewhat murky. They do offer very significant P-values for their volumetry in the methods.
Tractography data, showing islands of correlation with introspective ability. Graphs show island correlation (Aroc) and control graph of actual visual ability correlation (d').
The authors conclude that they have found some robust correlation between volume variations in a few brain areas and mental performance variations in this intriguing task of metacognition, or introspection. They further conclude that based on the past studies with other types of alterations and interventions, these correlations amount to causation, indicating that more brain matter and connectivity in Brodman's area 10 generates better self-knowledge in this simple type of visual task. They have no idea whether these variations arise from genes, developmental training, or recent training, the latter of which open some prospect for positive education.

In any case, aside from being another close correlation between brain and mind, the work is one more interesting instance of variation in human mental abilities, which on its own puts me in some awe of our diversity. If humans are each different in their very perceptions and subjectivity, what does it mean to be human? One reason the soul hypothesis is so attractive is that it would validate a sort of species-ist uniformity to the human condition.. that at the core we are all identical in the most important way.

But if we aren't.. if everyone is fundamentally different, then there is no such thing as the human condition and we have to negotiate in the open about our various beings and needs, without being able to assume a lot about inalienable rights, creator-given essences, "natural" behaviors, etc.

  • One human variant, with autism.
  • The Ur-blogger.
  • Skidelsky: "Tax persistent and excessive current account surpluses".
  • Mortgage and finance fraud, bigger than ever. Can we dream about resolving it?
  • The financial fraud documentary, Inside Job.
  • Bill Mitchell quote of the week, from "Why budget deficits drive private profit":
"The point is that when the government runs a surplus it reduces profits via its squeeze on aggregate income. That is why all the business sector should be screaming at the fiscal austerity plans that are rampant at present."
"So the fact that the business groups often lead the charge against budget deficits reflects the triumph of ideology over good judgement and the triumph of ignorance over understanding. I just shake my head in wonderment when I see a business person railing against budget deficits."

1 comment:

  1. I’ve always thought of myself (ha!) as an introspective person if it means really thinking about what I think and feel. As I recently wrote: “I trust no one in general, in particular, myself.” Some artificial propellants accelerate my thinking whereas others accelerate my emotional state. Alas, I have not found within me the ability to arrive at logical conclusions, (at least I know this) perhaps largely due to a lack of education, nor have I achieved such a profound unexplainable emotional union with the otherlies that causes me to (a) resort to blind faith as my reason or (b) insist that I have had a divine revelation. I am comfortable with “Could be wrong, could be right.” within the confines of my various thoughts or feelings.

    Sadly the high branch on the tree of knowledge is inaccessible to me at my age. On the other hand, life has afforded me a vast array of experiences from I have derived enough emotional content in which to drown. But even this is not enough. Therefore, I further explore my feelings with the imagination drawn out in me only by creative writing. I willfully place myself in scenarios that will afford me greater introspection. Often the conclusions I come to in my writings are simply those that any other man in a helpless state might arrive. And after such a laborious exercise, being satisfied with the process itself, I dismiss the results I have just declared. I think the term for this is heteropheonmenology, but I couldn’t swear to it. I would agree that the brain is both finite and accessible. But as for me – the me I am and the me that I want to be – well, that still feels both infinite and largely inaccessible. For all my introspection.