May 20 2008

More Sloppy Thinking from Michael Egnor on Neuroscience

Dr. Egnor, writing for the Discovery Institute blog, has some more tortured logic for us. He is desperately trying to defend his justification for dualism – the claim that the mind is more than the functioning of the brain, and like the creationists with whom he has found a home, he will not discard an argument simply because it is false.

In our latest exchange I first wrote that a recent bit of neuroscience research showed that brain activity comes before any evidence for mental activity. This is not an isolated result of this one study, but is generally what we find when we correlate brain function to mental function – the brain function comes first.

This point is relevant to the materialist-dualist debate. One of the lines of evidence for the materialist theory that the brain entirely causes the mind is that brain function correlates with mental function. The dualists employ a tactic of denial by arguing that this is merely correlation and therefore not evidence for causation, and that this correlation could result from the mind causing brain activity. While logically correct (correlation does not prove one specific causation), you can still infer causation from multiple correlations, and they ignore the many independent lines of evidence that suggest the arrow of causation is from the brain to the mind.I offered this study as just one such line of evidence. If the brain causes mind, then brain activity should precede mental activity. If the mind causes the brain, then it is reasonable to expect that mental activity could manifest either before or simultaneously with brain activity. This derives from the indisputable law of nature that in this universe causes must come before their effects.

Egnor has managed to misinterpret this simple line of reasoning now for the second time (not content, apparently, to have done so only once). He also reveals his broader strategy in arguing for dualism – interpreting the current technological and methodological limitations in neuroscience as evidence that something more than the brain is required to explain mental phenomena. Simply put – his position is an elaborate “god-of-the-gaps” argument for dualism.

For example, he writes:
Consider Dr. Novella’s own criterion for this debate:

“If the mind is completely a product of the material function of the brain, then we will be able to correlate brain activity with mental activity, no matter how we choose to look at it.”

Dualism predicts the inverse:

If dualism is true and the mind is partly the product of the material function of the brain and partly the product of something else, then we will not always be able to correlate brain activity with mental activity, no matter how we choose to look at it.

That is the core of his argument – a lack of perfect correlation between our measurement of brain activity and our measurement of mental activity.

He also states:
The materialist prediction is that the correlation between mind state and brain state must be 100%, minus experimental error. Dualism posits that the correspondence between mind states and brain states is not exact, because there are aspects of mind states that are not identical to brain states. Dualism predicts that the correlation is less than 100%, and that this lack of correlation cannot be explained away entirely as experimental error.

Notice that he throws in the caveat – “minus experimental error.” This gives the appearance that he understands there will be lack of 100% correlation due to experimental limitations. But the word “error” does not cover it, and he does not offer any guidance as to what he means by “error” and how much of an effect it might have. This is critical because it cuts to the very heart of Egnor’s argument.

When correlating brain activity and mental function we must consider how, exactly, we are measuring both. With regard to brain activity, we have several tools: electroencephalogram (EEG) measures electrical waves produces by brain activity; functional MRI scanning (fMRI) measures local blood flow in the brain, which enables us to infer brain activity as increased activity leads to increased blood flow; and positron emission tomography (PET) which uses a radioactive tracer to measure brain metabolism, usually of glucose – and again infers brain activity from increased metabolism.

What we do not have is a tool for directly imaging neuronal firing in a living person. We therefore do not have direct access to the brain function that actually correlates with mental function. The tools we do have are used to infer brain activity only. For each instrument we have to consider sensitivity, specificity, and resolution. Resolution is of particular importance. EEG has a very low resolution – with EEG we are only seeing the net electrical effect at the scalp of large groups of neurons firing together. At best the EEG tells us about overall brain activity, and can give us regional information at about the resolution of the brain lobes (left temporal lobe vs right parietal lobe, etc.).

PET scanning gives us higher resolution, and can look at the deep structures of the brain, which EEG is not good at. But still what we get are computer colored pictures of diffuse blobs of brain activity. fMRI is similar to PET but the resolution is a bit higher. The real advantage to fMRI over PET is that it gives real-time information. We can use fMRI to image brain activity while a subject is performing a task.

But the resolution is still far too low to image the actual neurons that are firing and producing any specific cognitive activity. To say that such a tool should correlate “100%” with mental activity is absurd. This is like using a medium-sized ground-based telescope to view the surface of Mars and declaring, based upon these fuzzy images, that there are no small rocks on the planet’s surface.

Dr. Egnor writes:

Its spatial resolution is excellent (3 to 6 millimeters), and it is so precise and so reliable that we use it to plan surgical resection of tumors and epileptic foci very close to critical brain regions.

This may be useful for surgical purposes, but is still far too low resolution to image neuronal activity directly. It is seeing only larger brain structures – like nuclei or collections of neurons that serve a particular function. Egnor here is confusing the utility of fMRI for gross surgical procedures with fMRI as a research tool to understand the subtlety of brain function.

Even more important to this discussion is the mental activity we are measuring? That’s right – we have to measure or record the mental activity some way also. There is no gold standard measurement, there is no way to know for sure what mental state or function a subject is experiencing. At best we can infer their mental state. This is done either by having them perform a task, subjecting them to a specific sensory stimulation, or having the subject simply report on their mental state. But we cannot read minds – we cannot “see” their mental state. Therefore, even though a subject may be performing a task, we won’t know what other mental activity might be going on at the same time. We also won’t know exactly how they are mentally performing the task as there may be multiple mental roots to the same end.

Ironically, Egnor seems to understand this as he uses this same line of argument for his own ends. Regarding the particular study we are discussing (and which I will get to below) Egnor writes:

The regions of the brain that were activated are known to be associated in some circumstances with planning, but the association is often nebulous, and activation of these regions provides little in the way of information about the content of the thought. Perhaps they were thinking, “What if I don’t feel like pressing the button at all,” or “what shall I get for lunch?” or “how much are they going to pay me for doing this?” The mere recording of generic brain activity during the study is meaningless, because the subjects were continuously conscious, and there’s no clear evidence, without hemispheric localization, to infer that the brain activity involved an unconscious decision about hand selection, which was the whole point of the study.

Therefore, when considering the implications of brain activity occurring prior to mental activity, Egnor is able to see the complexities of correlating presumed mental activity with fMRI activity. And yet he still argues that there should be a 100% correlation, except for experimental error. Well, would he consider this inability to know everything the subject is thinking experimental “error”?

Let’s take a look at the specific fMRI study that Egnor and I are discussing. Briefly, the researchers looked at brain function while asking subjects to choose one of two buttons to push. At the same time they were viewing a stream of letters, and they were instructed to indicate with each choice what letter was being displayed at the moment they decided which button to press. The purpose of this was to estimate when the subjects were conscious of their decision. They concluded from their data that subconscious brain activity preceded the conscious decision by about 7-10 seconds.

Specifically what the study found was this. With each button push there was activity in the corresponding primary motor cortex and the supplemental motor area (SMA). These are the parts of the brain that execute a voluntary motor action, so this was no surprise, and is consistent with prior studies. Also consistent with prior studies, activity in these brain regions occurred several hundred milliseconds prior to the motor action.

Therefore, with the more straightforward aspects of using fMRI and correlating with objectively observed activity (motor activity) correlation is strong (but not 100% – more below) and the brain activity always precedes the motor activity.

The researchers, however, wanted to know about the decision-making process, which they suspected occurred prior to executing the motor action in those parts of the brain responsible for higher cognitive function. What they found is that 7-10 seconds prior to when the subjects recorded they had made their decision there was activity in two other brain regions – the frontopolar cortex and then the precuneus in the medial parietal lobes. Further, which side (left or right) activity in these regions occurred was predictive of the ultimate choice about 60% of the time. While this may not seem impressive, these results were statistically significant.

How do we interpret all this? In my original post I had this to say:

However, they could only predict with 60% accuracy – random guessing being 50%. Given the trickiness of interpreting fMRI data and performing these kinds of studies, I am not at all impressed by the 60% figure. If it holds up to replication, and can be refined to be more accurate, then I would take it seriously. At this point I think any interpretation would be premature.

And yet Egnor is trying to characterize me as pinning my materialist hopes on this correlation. Let us take a closer look at this data. The researchers used fMRI with a computer program that used pattern recognition and statistical analysis to pull out those patterns of brain activity that seemed to predict the choice the subjects were going to make. The purpose of this was to determine which brain structures were likely the location of the decision making.

This is a very fuzzy analysis – which is why I gave it such a luke warm interpretation. To put this into perspective, the same analysis showed that the motor cortex correlated with side choice only 75% of the time. Since the motor cortex activity is directly tied to the motor action of pushing the button why wasn’t this correlation 100%? Well, that gets back to my point about the limitations of these techniques, including those points Egnor acknowledged himself. There are all kinds of activity occurring in the brain all the time. The primary method used in fMRI studies to isolate specific brain regions is to use some kind of statistical analysis – to show that activity in one particular part of the brain has a higher chance of occurring with a specific activity. Sometimes fMRI images are averaged out over many subjects – so the nice pictures represent a composite of many images, not a real image of one subject. That is basically what they did in this study – using statistical analysis across subjects to test the “predictive power” of various brain regions.

Egnor characterizes these results as “barely greater than chance,” but actually they were statistically significant – which by definition means they were significantly greater than chance. He is confusing accuracy with statistical significance. But this is a small point – I acknowledged that the small effect size weakens the interpretation of this one study.

About the 60% accuracy, Egnor writes:

If Dr. Novella is claiming that the experimental error is genuinely 40% (which it must be if materialism is true), what evidence can he provide that all of the lack of correlation is mere error? Can Dr. Novella cite any other fMRI studies that have demonstrated such astonishingly high experimental error? Does he even understand that there is a problem with his claims about the research?

Egnor is saying that only “experimental error” can explain the low correlation of localized brain activity to the subject’s choices. This is profoundly ignorant of the nature of fMRI research. The very nature of the fMRI technique in such studies is statistical. Looking at brain activity in real time with fMRI involves looking at a sea of noise and trying to pick out a signal peaking out above that noise. This is a fuzzy technology. That’s why the motor cortex activity only correlates 75% with motor activity. That is why I was so cautious in interpreting this one study, and why we need to correlate multiple studies to see if patterns of brain activity are consistent.

As a specific example from this study, the interpretation of this data is that the frontopolar cortex may be involved in planning a decision. The subjects knew they were going to make a decision, so subconsciously the decision was planned and then stored in the precuneous. Then they acted on their decision and pressed a button. However, it is obvious that it does not take humans 7-10 seconds to make a decision. That would give us dangerously slow reflexes. Therefore the researchers speculate that some subjects could have reversed their decision just before making the choice – they rejected the pre-choice they had made subconsciously. They may even do this at the last stage of activating their motor cortex, perhaps making a false start with one hand then switching to the other. This would partly explain the only 75% correlation with the motor cortex.

This is not “astonishingly high experimental error.” This is the nature of this type of research, which is very tricky because it must rely upon techniques that make it more likely for the subjects to be doing what you want them to be doing while looking at their brain activity.

Now let’s get back to materialism vs dualism. What Egnor is saying, which is truly ridiculous, is that the results of this study, and all the other fMRI studies of brain function, are evidence for dualism because of the imperfect correlation between mental activity and brain activity. His false premise for this conclusion is that fMRI’s are “so precise” only “astonishingly high experimental error” can explain such outcomes in a purely materialistic world. Therefore, he concludes, the most parsimonious interpretation is that something other than brain activity is at least partly responsible for the mind.

However, as I have now made clear, the sloppy correlation is easily explained as a limitation of the fMRI studies used to localize brain function.

I will also add that, as far as I am aware, Egnor has never proposed any scientific notion for what this extra-cortical mind-causing thing might be, how it functions, how it interacts with the brain and the physical world, or how we could measure it. His case is based upon entirely on imprecision in our current measuring technology and techniques. This is an argument from ignorance logical fallacy – a god-of-the-gaps strategy.

Egnor also does not present evidence for mental function in the absence of brain function. Nor has he presented any evidence for mental activity that does not correlate with brain activity at all, or within our ability to reliably measure it. He has not shown that the fMRI results are anomalous and therefore require a new phenomenon to explain. He has chosen for his criteria for evidence of dualism “lack of perfect correlation,” which gives him the opportunity to interpret experimental limitations as evidence for dualism.

If fMRI studies are so limited then why am I citing them as evidence for materialism? I offered this study as one line among many – not sufficient evidence by itself. I was pointing out one aspect of the data – that brain function always appears to precede the mental functions with which they correlate. fMRI studies are useful, but they are tricky and at present they need to be viewed in the proper context of their limitations. But taken together they are making progress in further exploring neuroanatomy and how it relates to mental function.

To end his article Egnor writes:

[Quoting me] “The problem with [dualism] is that it is unnecessary – it is adding an unnecessary step and violates Occam’s razor.”

William of Occam was a 14th Century English scholastic philosopher and a father of modern epistemology. He was also a Franciscan friar, and by his vows (God is Spirit, and man is created from dust and in God’s image)” he was a dualist.

So even dualists end up, posthumously and incongruously, in the materialist camp. Denialists are everywhere.

This is an excellent example of Egnor’s willingness to mangle logic in order to make what he thinks is a point. I never said that Occam was a materialist, nor did I cite Occam as any kind of authority on this specific question. I invoked the principle known as Occam’s razor – which states that when multiple interpretations of the data are possible, the one that introduces the fewest new assumptions is preferred. Stated another way – we should avoid unnecessarily multiplying assumptions.

Dualism introduces a new and completely unknown to science phenomenon of nature that is capable of producing mental effects but cannot be seen or measured in any way and for which there is no theory of mechanism. This is an enormous violation of Occam’s razor. Materialism builds upon what is known and established. We know that brain function is necessary for mental function and that it correlates with mental function within our current ability to measure it. A new phenomenon should only be added if it is necessary – and neither Egnor nor anyone else has established that it is necessary.

To recap – Egnor misrepresented what I actually wrote, he completely misinterpreted the fMRI literature in general and this study in particular, and he has relied entirely upon faulty logic to make his case.

And yet Egnor tries to characterize me thusly:

Dr. Novella is a materialist ideologue. He has difficulty drawing coherent scientific inferences, and his rhetorical style is little more than condescension and contempt. He brooks no questions. When it comes to challenges to his personal materialistic ideology, Dr. Novella sneers, dissembles, and ultimately invokes parsimony:…

Materialism within neuroscience is not my “personal ideology,” it is the consensus of scientific opinion. I also have no ideology – I am simply giving my analysis of the relevant logic and facts. I can be convinced of a different conclusion, if the facts warranted. He says I “brook no questions” – meanwhile I regularly do so in this very blog, as well as in my other blogs and my podcast.

But he did get one thing right – I do have condescension and contempt…for him. Egnor has chosen to enter the arena of public discourse specifically so that he can attack science and reason and promote his religious ideology. He uses his writings to confuse and misdirect, and to undermine the public understanding of science. He deserves all the criticism he gets. Rather than whining childishly about it (you may recall that Dr. Egnor appeared in the movie Expelled so that he would complain about people on the internet being mean to him), Egnor can either refrain from public discourse, or more carefully examine his own logic.

14 responses so far