Here is an excerpt from Alfred R. Mele, Free Will: Action Theory Meets Neuroscience:
In a recent article, Libet writes: "it is only the final ‘act now’ process that produces the voluntary act. That ‘act now’ process begins in the brain about 550 msec before the act, and it begins unconsciously" (2001, p. 61).10 "There is," he says, "an unconscious gap of about 400 msec between the onset of the cerebral process and when the person becomes consciously aware of his/her decision or wish or intention to act." (Incidentally, a page later, he identifies what the agent becomes aware of as "the intention/wish/urge to act" [p. 62].) Libet adds: "If the ‘act now’ process is initiated unconsciously, then conscious free will is not doing it."
I have already explained that Libet has not shown that a decision to flex is made or an intention to flex acquired at -550 ms. But even if the intention emerges much later, that is compatible with an "act now" process having begun at -550 ms. One might say that "the ‘act now’ process" in Libet’s spontaneous subjects begins with the formation or acquisition of a proximal intention to flex, much closer to the onset of muscle motion than -550 ms, or that it begins earlier, with the beginning of a process that issues in the intention.11 We can be flexible about that (just as we can be flexible about whether the process of my baking my frozen pizza began when I turned my oven on to pre-heat it, when I opened the oven door five minutes later to put the pizza in, when I placed the pizza on the center rack, or at some other time). Suppose we say that "the ‘act now’ process" begins with the unconscious emergence of an urge to flex – or with a pretty reliable relatively proximal causal contributor to urges to flex – at about -550 ms and that the urge plays a significant role in producing a proximal intention to flex many milliseconds later. We can then agree with Libet that, given that the "process is initiated unconsciously, . . . conscious free will is not doing it" – that is, is not initiating "the ‘act now’ process." But who would have thought that conscious free will has the job of producing urges? In the philosophical literature, free will’s primary locus of operation is typically identified as deciding (or choosing); and for all Libet has shown, if his subjects decide (or choose) to flex "now," they do so consciously.
What Libet et al. want to show is that the notion that conscious willing plays a genuine role in the etiology of a behavior such as flexing a finger is illusory. Their evidence for this is that the process in the brain that initiates the action begins some 550 milliseconds before the action and is unconscious. Only 400 msecs later does the subject become aware of his wish or urge or intention or decision to act. This is supposed to show that the conscious intention is not causally efficacious and that conscious will is an illusion.
Mele rebuts this argument by showing that it trades on a confusion of decisions/intentions on the one hand and wishes and urges on the other. To want to do X is not the same as to decide to do X. Phil may want another Fat Tire Ale but decide not to drink another because he has already decimated Bill's supply and doesn't want to presume on his host. So even if the wanting to do action A begins in the brain a half a second before the doing of A, and is unconscious, it doesn't follow that the decision to do A begins in the brain a half second before the doing of A and is unconscious. Free will is displayed in decisions and choosings, not in wants and urges.
Basically, what Mele does quite skillfully in this article is show the indispensability of accurate conceptual analysis and phenomenology for the proper interpretation of empirical findings. The real illusion here is the supposition that the empirical findings of neuroscience can by themselves shed any light.