The following analysis was submitted by Brains reader Bill Skaggs, a neuroscientist whose published work is in electrophysiology, but who has been working for some time on a book about the relationship between consciousness and the brain. He blogs at https://weskaggs.net. — JS
I would like to examine a famous set of experiments carried out in the 1970s by Benjamin Libet. The results that he obtained are widely considered paradoxical, but what I would like to argue is that they are really only paradoxical if interpreted dualistically. When properly understood, they function as a compelling refutation of dualism — even though Libet himself did not understand them that way.
The aim of Libet’s experiment was to gain an understanding of how free will manifests itself inside the brain. Libet’s goal was to design an experiment in which human subjects would choose, of their own free will, the time at which to execute a certain action. He wanted to record their brain electrical activity using an EEG system, and look for changes in relation to the time at which the decision to act was made. Specifically, he decided to focus on the so-called “readiness potential”, a large EEG wave that typically begins several hundred milliseconds before a person executes a motor response such a pressing a button.
Here is the methodological problem that Libet faced: it is easy to measure the time at which a subject presses a button, but how can you measure the time at which a subject has made a decision? The approach that Libet took was quite innovative. His subjects were instructed that at some time during a long interval they should press a button, and that they were completely free to choose the moment to push it — in fact they should try to do it in as random a way as possible, not as a result of any sort of cuing. He also asked his subjects to watch a bright dot of light that swept around a circle on an oscilloscope screen, and to note its angular position at the moment they decided to act. This information allowed the “decision time” to be calculated with a precision of slightly better than one twentieth of a second.
And here is what Libet observed. The onset of the readiness potential could be seen in the EEG several hundred milliseconds before the calculated decision time, that is, several hundred milliseconds before the time calculated on the basis of the reported dot position.
It is important to be clear that everything I have said up to this point concerns data, not inference or interpretation. Each trial of the experiment generated three time measurements:
- The time at which the button was pressed, recorded mechanically.
- The reported time of decision, calculated from the oscilloscope angle reported by the subject.
- The time of onset of the readiness potential, calculated from the recorded EEG signal.
The result that Libet obtained was that the time of onset of the readiness potential was systematically several hundred milliseconds (that is, several tenths of a second) earlier than the reported time of decision. In other words, physiological evidence of the decision appeared well before the moment at which the subjects reported that they made the decision.
The question is, how should that result be understood?
To many people, Libet’s finding seems counterintuitive, or even impossible. How could physical traces of the decision appear before the decision had been made? The only way to make sense of the results, it seems to many people, is to conclude that the decision was not really made when the subject believed it was made — that the sense of having made a choice at that moment was an illusion. Either that, or something must be wrong with the experiment methodologically.
Indeed, it is not all that difficult to find problems with Libet’s methodology. From an experimental scientist’s point of view, his method of identifying the decision time is problematic. Normally in an experiment it is desirable for the experimenter to have control over the independent variable. In this case, though, the experimenter has no control over the time at which the decision is made, and therefore can’t rule out the presence of unobserved variables that may influence the decision. Other concerns have also been raised about the validity of the subject’s report as a measure of the decision time.
Really, though, that doesn’t actually matter very much. The aspects of the Libet experiment that make it philosophically important are immune to those sorts of objections. Here is what makes the experiment important:
- The result is impossible to make sense of in the framework of dualism.
- Outside the framework of dualism, the result is not only easy to understand, it is inevitable, at least at a qualitative level.
Let me first spell out why the result is impossible to understand in the framework of dualism. If we view the Libet paradigm in terms of a machine-and-operator metaphor, we will see the decision as taken by the operator (by “consciousness”, as many people put it), and the EEG signal as something that happens in the machine (the brain). Libet’s result, then, means that signs of the decision can be detected in the machine before the operator has actually made the decision. And that, of course, seems paradoxical. The only way to interpret it (within the framework of dualism) is that the operator is only under an illusion of having made the decision, and that the decision was really made by the machine. This is in fact the most common interpretation of Libet’s experiment, and many people (including Libet himself) have seen it as the only possible interpretation.
But let’s now look at what happens if we resist the temptation to impose a dualistic interpretation on the data. Suppose we look at the scenario simply in terms of brain activity and its relation to behavior. The crucial thing is the neural process by means of which the subject takes note of the decision time. This must, at some level, involve the execution of a brain activity sequence that causes an observation of the dot position to be made and the result to be encoded within short-term memory. Libet’s result shows that this observing-behavior is systematically preceded by brain activity of some sort. But isn’t that inevitable? Would it be credible for this brain activity to execute with nothing whatsoever coming before it in order to trigger it? That would be a miracle. The precise timing observed by Libet is not inevitable, of course, but the fact of something occurring in the brain to cause the making of the decision cannot be avoided.
Thus when we look at the data without forcing it into a dualistic framework, we see that it is not paradoxical at all.
The real importance of Libet’s experiment is that it is a compelling refutation of dualism. From a non-dualistic point of view the result is not problematic in any way, but from a dualistic point of view it is impossible to make sense of.
It is worth noting, ironically, that Libet, although his historical importance derives from having refuted dualism in this way, was himself a thoroughgoing dualist. He was what Daniel Dennett called a “Cartesian materialist”. Libet wrote many papers about his experiment and variations on it, but all of them viewed the data within the context of a machine-and-operator metaphor. (He used “consciousness” as his name for the operator.) He thought that his experiment proved that consciousness could not initiate behavior; however he also thought that it might be possible for consciousness to veto behavior in some cases.
The conclusion that is most commonly drawn from the Libet experiment is that free will is an illusion — that we do not really control our own behavior — that the actions we believe are freely willed are actually generated by unconscious processes occurring inside our brains. I don’t believe that this is the correct conclusion. The correct conclusion, I believe, is that the intuitive concept of free will is intrinsically dualistic, and that any attempt to apply it in a non-dualistic framework will result in paradox.
(Footnote: I make no claim to be the only person who has ever figured this out. Libet published a “target article” on his work in Behavioral and Brain Sciences that was followed by several dozen responses, and some of the commentators clearly grasped the underlying dualism in Libet’s way of thinking. But none of them really explained it very clearly.)
Great post! But do you have any thoughts on “but the fact of *something* occurring in the brain *to cause the making* of the decision cannot be avoided”? Much appreciated.
As a preface let me quickly note that I didn’t realize this was going to be posted yesterday and was away from my computer all day — I’d like to thank all the people who commented, and I’ll try to respond to all the points that have been raised.
Thomas, in regard to your question, the “decision making” circuitry in the brain is incredibly complex, and we are hardly beginning to get a handle on it. At the moment it’s hard to say more than that the basal ganglia and prefrontal cortex play essential roles. There are a few computational models of neural mechanisms for action selection, but they are very primitive.
Regards, Bill
“The result is impossible to make sense of in the framework of dualism.”
‘Impossible’ is a very strong claim. I might agree that the results seem to cohere better or be more reasonable with a neural viewpoint, but the dualist can at least explain how it is *possible*.
My hunch is that Libet doesn’t pose any unique problems for dualists over all the problems dualists already face,.
For instance, substance dualists have a lot of options they could appeal to. E.g., they should all say they realize the brain is extremely important for generating experiences, that there is a strong coupling between mental states and brain states. For instance, we (speaking a as a substance dualist) think that the (unconscious) processing going on in V1 is what largely determines the visual experiences of our nonphysical mind.
As for Libet, we are fine with the claim that, even with something like decision making, there are complex unconscious neuronal processes that our nonphysical mind doesn’t control or have access to, and which strongly shape behavior and consciousness. This is no more mysterious than our view on visual perception, that it is also determined by the information processing going on in V1.
They don’t need to believe that the consciousness of intention to do X is what causes us to do X. Whatever story the neuroscientists give, they can give the same story, but with a dualist twist. At least right now, anyway.
Of course you could ask how these two different types of processes are supposed to interact, etc, but that is a problem they face independently of Libet.
Property dualists should be able to handle Libet relatively easily. Since they tend to think pretty much everything is conscious, you could ask them why we are not conscious of more events going on in our brains, like the informational side of the readiness potential leading up to the behavior. But again, that is a generic problem they have that Libet doesn’t really uniquely pose IMO. I think they can overcome such generic problems with some creative distinction-making.
Eric, thanks for the comments. My post was extracted from a larger context, and there is an aspect of that context that I unfortunately did not make clear enough. What I mean by “dualism” here is what Churchland called “popular dualism” and I call “metaphorical dualism” — Ryle’s “ghost in the machine”, i.e., having a mental image of the brain as a machine and the mind as its operator. When I wrote that the Libet experiment is impossible to understand in the framework of dualism, what I was trying to say is that it cannot be understood using the “ghost in the machine” metaphor. I would not claim that it is impossible to reconcile with property dualism or substance dualism, though I think that for substance dualism it would be difficult.
The relationship between property dualism, substance dualism, and popular/metaphorical dualism is rather complex, and I don’t have space here to explore it. I have written out my views at length in an essay called “Metaphorical dualism and the Cartesian Airplane”, which can be found on my web site at https://weskaggs.net/?p=33.
More generally, I would be wary of ever saying dualism has been empirically refuted (especially property dualism). You may underestimate the ingenuity of some dualists. 🙂
Per my reply above, I agree entirely, and I also agree with the comment by Richard immediately below this one. In fact I would go so far as to say that most forms of property dualism violate Occam’s razor and therefore cannot even in principle be refuted scientifically.
Regards, Bill
To add to Eric’s point, it’s worth noting that epiphenomenalists, in particular, are going to embrace the exact same causal story (for explaining physical events and behaviour) as physicalists do.
Mr. Skaggs, regarding my initial comment, I suppose what I find bothersome about this experiment, as you’ve described it, is first the ambiguity of terms. For example, pressing a button is seen as a “report” of a unconscious decision rather than the “implementation” of a conscious decision. I don’t see this as a clear distinction between what is “unconscious” or “conscious.” Also, given the methodology it seems to address the issue of decision-making in a very trivial sense. In other words, even though randomness is encouraged in the button-pressing, the subjects are essentially only following some simple instructions. But is this what a reasonable person thinks of as “decision-making.” When I yawn, I find it hard to believe that I’ve made a decision to yawn, even though it may be reported that I yawned.
Anyway, being a non-dualist, I find your conclusion “that the intuitive concept of free will is intrinsically dualistic” to be compelling.
Thomas, I agree with you to the extent that I would be leery of using that terminology in a paper of my own. But that is the terminology that was used in Libet’s papers and the literature that followed them, and if I am going to discuss that literature, I don’t have the option of avoiding that terminology.
Regarding the question of whether yawning is a “decision”, that’s a very thorny issue that many people have been working on. It seems clear that the intuitive distinction between voluntary and involuntary behaviors corresponds to *some* distinction at the neural level, but theoretical accounts are very confused right now.
Regards, Bill
This is an aside question:
does anyone know of arguments concluding that we should accept Event Readiness Potentials (ERPs) or bereichshaft potentials as either (i) indicative of a decision or (ii) identical to a decision?
I once found a old paper in German this topic (but I don’t know German).
It seems that this auxiliary assumption is necessary for a wide range of experiments, so I want to know why we are taking Libet-style experiments to be falsifying things other than this assumption that ERPs represent decisions (somehow). After all, the claim that Libet falsifies this assumption is logically possible.
Any guidance would be thoroughly appreciated!
Nick, regarding the relevance to the Libet experiment, the only claim that is necessary is that the ERP cannot appear before the decision has been made. Since it contains information about the decision, that conclusion seems unavoidable, if the word “decision” has any meaning at all here.
To say that the ERP is identical to the decision would be like saying that the roar of the crowd in a football stadium is identical to a scoring play. That just doesn’t work. The idea that it is an indicator makes more sense (just as it makes sense to say that the roar of the crowd is an indicator of a scoring play.)
Regards, Bill
Right. Thanks for replying.
So what reason do we have to think that the ERP is to the decision as the cheering in the stadium is to scoring. Is it that the ERP always coincides with a self-reported decision? Is it that the ERP always councides with button pushing?
My personal view is that a “decision” is an interface-level entity, which can’t be identified with any specific event occurring inside the brain. Trying to identify decisions with brain events leads to the sort of difficulties that Daniel Dennett spelled out in chapter 5 of *Consciousness Explained*. (He was talking about perception rather than action, but the logic still applies.)
In any case, my understanding is that a readiness potential can often be seen for actions that are contemplated but never actually performed
Thanks Bill. The reference to Dennet’s chapter is particularly helpful. Your patience with my question is thoroughly appreciated! Again, great post!
Mr. Skaggs, thank you for replying to my comments. I plainly lack the learning and intellectual sophistication that are needed frame my concerns, but it is nevertheless an area of interest to me. I try to pick up what I can given my limitations. I have gone to your site to read some of your articles there. I want to compliment you on your skills as a writer considering the nature of the subject matter. Again, thank you for taking the time to reply.
Thanks. If you want to explore this topic more deeply, let me recommend these books:
1) “The Illusion of Conscious Will”, by Daniel Wegner
2) “Your Brain is (Almost) Perfect: How We Make Decisions”, by Read Montague
3) “The Mind Within the Brain: How We Make Decisions and How Those Decisions Go Wrong”, by A. David Redish
I hate the title of the Montague book, but it’s well written and has a lot of good information. Wegner is not such a good writer and sometimes gets tangled up with dualism, but his book is still quite useful. The Redish book, which just came out a couple of months ago, covers the neuroscience of decision-making in a very readable but also authoritative way.
Regards, Bill
As an aside, the original title of Montague’s book is “Why Choose this Book?”–I have a version with that title. I think that the publishers realized that wasn’t a great promotional idea. Hence the new title (same book). I also dislike the new title, though I love the book.
Thanks for the posts!
“in fact they should try to do it in as random a way as possible”
This suggests to me that the subjects have been explicitly instructed to *not* make a conscious decision.
The experiment starts with the assumption (correct I think) that cued (determined) decisions are not free, so tries to encourage them to be random (indetermined) instead. Yet even if a decision could be a truly random, then I don’t see how that would be free.
This is just a mirror of the whole determinism vs indeterminism vs where to shoehorn-in free will debate, and regarded this way the experiment appears designed to miss the target.
I don’t think anyone, dualist or otherwise, would claim that *every* human action is the result of free will. For me, free will is something we use to intercede in our otherwise (faux-)determinate decision-making.
True randomness is not possible. What doing the task “in as random a way as possible” actually involves is *resisting* habits and internal/external mental cues that might trigger us to press to the button.
In this case at least, it’s there (if anywhere) that you’ll find free will in action.
Hi Matt.
Let me start by citing the best source for Libet’s views and critical reactions to them: LIBET, B. “Unconscious cerebral initiative and the role of conscious will in voluntary action.” Behavioral and brain sciences 8.4 (1985): 529-566. Two of the commentaries there, “Libet’s dualism” by R. J. Nelson, and particularly “Pardon, your dualism is showing” by Charles C. Wood, relate closely to the theme of my original post.
In that BBS article Libet wrote: “The subject is
instructed to allow each such act to arise ‘spontaneously,’ without deliberately planning or paying attention to the ‘prospect’ of acting in advance. The subjects did indeed report that the inclination for each act appeared spontaneously (‘out of nowhere’), that they were consciously
aware of their urge or decision to act before each
act, that they felt in conscious control of whether or not to act, and that they felt no external or psychological pressures that affected the time when they decided to act”. Whether that setup makes for a “conscious decision” is perhaps not 100% clear, but it’s not easy to see how he could have done better.
Other than that, I’d just like to note that your own definition of free will is implicitly dualistic. When you say “free will is something we use to intercede in our otherwise (faux-)determinate decision-making”, it’s clear that you are visualizing “we” as the operator and “our decision-making” as events that take place in the machine. (If you reject that, I’d like to know what you intend the word “we” to refer to.)
Best regards, Bill
Hi Bill,
Thanks for the response.
Yes, it’s exactly that “out of nowhere” decision making that I’d argue isn’t a product of free will. Free will for me (if it exists at all) is somewhere between the unforced instinct to act and the act itself, and is (at least) sometimes preceded by conscious deliberation.
So for me, it would have been better to look at both the type of response studied here, along with the one I’ve described (and any others we can think of!) and look at the differences between them. Of course, I have no idea how viable that is!
In regards to my definition of free will as being dualistic, I’m struggling to make sense of that as I just think of it as a causal process that may or may not exist. But assuming you mean that my definition of consciousness is that, yes and no and not sure.
I wouldn’t subscribe to substance dualism, but am open to property dualism. My own current take (i.e. pet theory) is monistic I think, as the only dualism is one of scale: some sub-Planckian states of the universe instantiat matter and energy at the super-Planckian scale and some don’t. Experiential consciousness does not.
However, pet theories aside, I’m not sure it matters to free will how one’s take on consciousness goes. Why couldn’t it exist under a physicalist view?
Matt
I pointed in an earlier comment to an essay on my web site that I wrote in an effort to clarify these issues, but to answer briefly, one of the most useful ways to recognize implicit dualism is to apply the “robot test” — to ask yourself whether a definition could possibly be applied to a robot. Would it be possible to have a robot that could intercede in its own decision-making? If so, how could that be recognized? And if so, would you say that such a robot has free will?
Regards, Bill
The robot criteria seems to presuppose a computational theory of consciousness rather than simply monism, no?
Or we have to assume that ‘robot’ holds for any kind of organised physical system, which is very counterintuitive (‘animals are robots’ would almost be a tautology).
Yes apologies, I will take a look when I’m able and perhaps reply more fully. On first sight my answer would be:
– yes, absolutely a robot could intercede in its own mechanical decision making, if and only if from (presumably) its hardware or software emerged experiential consciousness, and it’s cognitive abilities were as sophisticated as our own.
– as far as we know so far, it could not be recognized from the outside. However, the robot would believe itself to be conscious and perhaps to have have free will and could report that belief. If we had not programmed it to report that, this would be good evidence. (exactly the same situation as with ourselves)
– I would accept that both its reported experiential consciousness and free will may be illusions, but would follow Carl Sagan’s suggestion that extraordinary claims need extraordinary evidence. Therefore, barring such evidence (it would have to be a great deal better than Libet’s) I would assume the situation to be as universally reported by it and several billion other similar systems (humans) on the planet, and would look to every possible scientific and metaphysical avenue to explain that phenomena.
(This is a reply to quen_tin)
The robot test doesn’t really presuppose any sort of theory — it’s a trick to reveal the mental picture that a person is using. Most people visualize the mind-brain interaction using a “ghost in the machine” metaphor (as Ryle put it); but most people visualize a robot as a machine that doesn’t contain a “ghost”. The consequence is that if a concept implicitly presumes a “ghost in the machine” framework, most people will not see how to apply it to a robot. You can see this in action when people ask questions such as, “Do I have free will, or am I just a robot?” There is no deep philosophical significance here — this is just a device for bringing covert thought processes into the foreground.
(The idea is originally due to Dennett, by the way.)
Regards, Bill
I understand but I think (contra Dennett perhaps) that the trick does not reveal necessarily a dualist framework but a non-computationalist framework.
To be precise I have the intuition that something which does not seem to apply to a robot can be physical nonetheless but maybe we should take examples to see how the trick works in practice.
Anyway thank you for the article and the comments which are very illuminating!
As an example, many people apparently think it is meaningful to say that people are the owners of their thoughts. But what happens if you try to apply that concept to a robot? Would anybody ever say that a robot is the owner of its thoughts (or its computations, if you prefer)?
Good example indeed. That’s food for my thoughts 😉