What I'm going to do is to describe three problems that all three experiments run into in moving from the data that they generate to the conclusion that there's no free will. Again, I won't be doing anything fancy philosophically. I'm going to concentrate on the data and what the data shows. So, here's problem one, the first problem shared by all three, but we start with the Libet study. So, recall that Libet thinks something very special happens at minus 550 milliseconds. That is a little more than half a second before muscle motion begins. He thinks that the special thing is that our decision is made then. Now if that's right, one question we should ask is, how long does it take a decision to do something now and in particular to flex your wrist now to generate muscle motion? If Libet is right, it takes a little more than a half a second. But is that true? What we should do then is look for separate evidence that would answer that question, how long does it take a decision to do something now to generate muscle motion when the muscle motion is wrist motion and you're flexing your wrist? Now, one way to get evidence is to look for reaction time studies in which subjects are watching one of these Libet clocks, one of these fast clocks. The reason you want them to be watching the clock is that watching the clock is going to divide attention between that and the other task you're doing, which is the reaction time task. Now, in reaction time studies, subjects know what they're supposed to do when they get a go signal, and what they're supposed to do is do that thing as soon as they can in response to the go signal. So, for example, in a reaction time study in which the go signal is a tone, and what you're supposed to do is flex your wrist, your aim then would be to flex your wrist as soon as you can in response to the tone. So, what I did, this was years ago now, I'd looked for a reaction time study in which subjects were using a Libet clock. They were watching a Libet clock. I found one, and it was a button pressing. So, what subjects were supposed to do when they heard the tone was to click a button as soon as they could. The meantime between the sounding of the tone and muscle motion was 231 milliseconds. So, now picture it this way. Suppose that what's going on is, detection of the tone generates a proximal intention that is an intention to click right now. So you detect the tone that generates the intention, and the intention generates a muscle motion. So, if the meantime between the sounding of the tone and muscle motion is 231 milliseconds, the meantime between intention and muscle motion is going to be a bit less than that because it's a causal process of detecting the tone. It'll take a little while to generate the intention. But now we're nowhere close to the 550 millisecond figure, and we are very close to the w time, the 200 millisecond figure. Now, when subjects aren't watching a clock and they're doing a similar reaction time test, reaction times tend to be shorter. In a study where subjects were flexing one wrist in response to a go signal, the meantime between go signal and muscle motion was 125 milliseconds. In the study where they're flexing two wrists, both wrists, in response to a go signal, the reaction time was even less. It was about 95 milliseconds. So what we have here then is evidence that what Libet is detecting at minus 550 milliseconds isn't yet a decision or intention to do a thing, it's something else, because if it were a decision or intention to do it right then, you'd get the movement much sooner than you do. So, now we move to the Soon study, and we ask what's going on at 7-10 seconds in advance of muscle motion or in advance of a button press. Now, I mentioned that the scientists reported that they were able to predict with high accuracy, that was the quotation, 7-10 seconds in advance which button the person would press next. The high accuracy was actually 60 percent. So, 60 percent accuracy. Now, just to give you a feel for what 60 percent accuracy is, notice that with a fair coin, I can predict with 50 percent accuracy which of two buttons you're going to press next. So, the scientists would do 10 points better in terms of accuracy, but I'm relatively close, 10 points lower. Also, I can do better than make these predictions 7-10 seconds in advance. If you promise me that you won't press a button for a minute, then with my fair coin, I can predict a minute in advance which button you're going to press next with 50 percent accuracy. So, the 60 percent accuracy figure isn't enormously impressive, and I suspect that what's being detected is a slight bias toward one button or the other and the next press. This is what I have in mind. What subjects need to do, what they're told to do, is not to fall into any patterns. So, what they do then is they have to keep track at least in a rough and ready way of which button they've been pressing lately, what their patterns might have been and try not to fall into any further patterns. I believe that what happens then is that what's generated is a slight inclination for the left button say or the right button that maybe the subjects are not conscious of. But if that's all it is, it isn't a decision or an intention. It's just a nudge. They're being nudged unconsciously in one direction or another but not pushed all the way. So, then, in the Fried and co-authors' study done with the depth electrodes, what's going on 800 milliseconds before muscle motion? Well, here too, I think that what's happening is that a process is up and running. Eventually, it will result in muscle motion because what the person is supposed to do is click sooner or later. But when in the process is the actual decision to press now made? If we look at the reaction time studies, it's going to be close to 200 milliseconds before muscle motion. I think the probability of raising or clicking soon rises as time goes by, but it doesn't rise all the way to one. As far as we can tell, the evidence is just an 80 percent chance of muscle motions soon. So, I think we don't have good evidence that decisions are made early, and the best evidence we have is putting the decision-making much closer to the time of muscle motion somewhere around 200 milliseconds before it. So, that is one problem that all three studies have in common, and short way to think about the problem is they haven't shown that decisions are made at these early times. The data are consistent with the decisions being made much closer to the time of motion and within that window of awareness time. So, it's quite possible that the decisions aren't made until people are aware of making them.
