In “Agent Causation” Timothy O’Connor makes a passing assertion that there are many unresolved questions for materialist agency as he posits it, and that many of these questions are empirical in nature and can only be resolved with “extensive advancements within neurobiological science.” [1] Two particularly salient questions are (1) “Precisely to what extent is an ordinary human’s behavior directly regulated by the agent himself, and to what extent is it controlled by microdeterministic processes?”[2] And (2) whether microdeterministic processes can be predicted or not. While O’Connor may believe that advances in neuroscience will reinforce rather than call into question his theory, this is not the case. Stretching from the 1980s to a recent study in 2008, neuroscience has demonstrated that predictive brain activity can be seen to occur prior to a test subject’s consciousness of making a decision. From Libet to present, these studies provide damaging replies to the questions which O’Connor’s theory leaves unanswered.
In 1985, Benjamin Libet crafted an experiment to measure what he deemed “unconscious readiness potential” (RP) and a test subject’s experience of having an intention or volition toward a specific, repetitive movement at chosen intervals. Libet measures Readiness Potential as voltage levels in the brain. Paling in comparison to today’s standards of FMRI machines and 3 dimensional mapping, Readiness Potential simply indicated an increase, decrease, or maintenance of brain activity as measured by changes in voltage. The original test was set up where brain activity was monitored as the subject performed the task, which allowed Libet to measure the interval between the beginning of brain activity, the claim of conscious feelings of volition, and the action being performed. The tests demonstrated a marked and measurable increase in brain activity prior to the subject reporting becoming conscious of deciding to flex his or her wrist. On average, 200ms elapsed between the subject becoming aware of willing the action and that action taking place. However, a change in the subject’s RP occurred 550ms prior to the actual flexing of the wrist. This meant that, on average, 350ms elapsed between increased RP in the motor cortex and the subject becoming aware of willing the action. In effect, measurable brain activity was occurring prior to the subject’s consciousness. Libet’s reading of these results was that, should it be the case that we choose to act out of consciousness, there would not be a 350ms period of activity prior to the subject becoming aware of the intention to act.[3] This would seemingly point to the conclusion that brain activity prior to consciousness, not conscious decision-making, was the cause of a subject’s action. At the very least, it would seem that such antecedent brain activity must be explained.
Adding a layer of complexity to Libet’s study were instances of ‘veto’ where subjects were capable of changing their decision to flex their wrists at the last moment, reporting that this occurred in the last 150ms of the experiment. Libet indicated it was this narrow margin which seemed to give hope that human action could indeed be caused, or at least effected by, conscious choice.[4] Discussing the case of veto scenarios, Libet writes that “The concept of conscious veto or blockade of the motor performance of certain intentions to act is in general accord with certain … views of ethical behaviour…‘Self control’ of the acting out of one’s intentions is commonly advocated.”[5] Such vetoes, however, are hardly the substantial platform upon which free will might rest that Libet believed them to be. As John Ostrowick points out in “The Time Experiments of Libet and Grey Walter,” it would seem that veto scenarios only offer a negative definition of freedom – the ability to stop oneself from acting rather than be the source of one’s actions. This, of course, is presupposing that veto decisions do not themselves possess unconscious antecedent brain activity.[6] In The Actor’s Brain, Sean Spence also argues that we would have good reason for postulating that all conscious vetoes are preceded by unconscious brain activity and that the act of consciously vetoing is merely a construct after the fact.[7]
Since Libet does not seem to offer any substantial reason as to why vetoes do not have similar antecedents, or is vague in that if antecedents exist they are unconscious, we are still left without a positive account of conscious deliberation or choice playing a causal role in action a la O’Connor’s defense of indeterministic reasons explanation.[8] I need not offer a detailed account for why O’Connor and other agency theorists must be committed to the claim that deliberation, consciousness, or reasons play some role in free action – this is because the agent’s ability to participate in action is an implicit requirement of any ‘interesting’ definition of freedom. For if reasons, deliberation, or consciousness play no role in action, then in what meaningful sense can the agent participate in her own actions? Alternately, if the cause of action is unconscious brain activity, conscious deliberations seems to have been preceded, and therefore superseded as a cause, by brain activity outside the control of the agent.
However, the details and science of Libet’s studies do leave room for error, as well as questions about the equipment utilized, etc. The margin of error for Libet’s experiment is quite small, measured in milliseconds, making it difficult to ascertain whether the margin of error was to blame for the conclusion that brain activity geared toward action occurs prior to a person choosing to act. Similarly, because RP measurements only demonstrate an increase in voltage in the brain, an argument could be made that such antecedent unconscious brain activity is merely preparing the agent’s brain for deliberation, choice, or conscious action. Recently, however, a new study has emerged that confirms Libet’s findings and provides substantial evidence that brain activity precedes conscious volition by a much larger degree than in Libet’s experiment and that this very same brain activity can be utilized to predict between two choices what choice the subject will choose with a significant level of accuracy.
Published in Nature Neuroscience, “Unconscious Determinants of Free Decision in the Human Brain” found that “the outcome of a decision can be encoded in brain activity of prefrontal and parietal cortex up to 10s before it enters awareness.”[9] Chun Siong Soon and the other authors of the study postulate that the delay between brain activity and the subject’s awareness is due the communication among high-level control centers preparing for action. This study made use of fMRI that monitored the subject as she viewed a series of letters on a screen. When the subject felt the desire, she pressed a button in either the left hand or right hand. After pressing the button of her choice, the subject then was presented with a selection of letters and asked which letter was being displayed on the screen when she felt the desire to press a specific button. In order to clarify the testing procedure, Figure 1 from the study is represented below.[10]
This study clearly surpasses Libet’s experiment in a number of ways. First, the technology utilized far surpasses the EEGs and oscilloscope used by Libet in its ability to record information but also in its ability to differentiate areas of the brain. For example, Libet tested only the motor cortex. This most recent study was able to simultaneously monitor 7 different areas of the brain and measure each area’s separate predictive capability. Secondly, as mentioned earlier, this study turned the several hundred millisecond interval between brain activity and consciousness into a 7-10 second interval, thanks in part to advances in technology. Thirdly, the ability to utilize brain activity to predict what choice the subject will take anywhere from 7-10 seconds prior to the subject even having the experience of deciding to choose:
Taken together, two specific regions in the frontal and parietal cortex of the human brain had considerable information that predicted the outcome of a motor decision the subject had not yet consciously made. This suggests that when the subject’s decision reached awareness it had been influenced by unconscious brain activity for up to 10 s, which also provides a potential cortical origin for unconscious changes in skin conductance preceding risky decisions.[11]
The predictive capability of this study is where, in my opinion, true ground has been gained toward demonstrating physicalist reduction claims. The Soon study demonstrates that the frontopolar and precuneus regions of the brain demonstrate the highest predictive capability prior to conscious decision, hovering around 59-60%. There is also a noted difference in predictive capacity between brain areas both before and after conscious decision. So, prior to the conscious decision the frontopolar and precuneus regions have a 60% predictive capacity and the motor cortices have predictive capacities at or below 50%. After the conscious decision, however, the cortices display a predictive capacity around 62% and the frontopolar and precuneus regions show diminished capacity at round 54%.[12] Even though Libet’s original experiment entailed the possibility that brain states might determine a specific action, the lack of predictability (paired with his margin of error), meant that his findings could be shrugged off as inconclusive, or just seen as proof that brain activity prepares a person to make a decision but does not determine which decision a person makes.
With the discovery of predictability comes the stronger argument that brain activity does not merely prepare for some sort of agent causal power but rather determines such behavior, all the while unbeknownst to the agent. As the Soon study says, “Also, in contrast with most previous studies, the preparatory time period reveals that this prior activity is not an unspecific preparation of a response. Instead, it specifically encodes how a subject is going to decide.”[13] However, there are two potential criticisms that might threaten physicalist conclusions, and represent potential replies from those who adopt O’Connor’s theory of agent causation, or those who may be skeptical of these conclusions.
Potential Criticisms
The first of these potential criticisms is that brain activity might encode meaningless choices, like when to press a button (and which button to press) or when to wiggle a finger. However, there is no guarantee that complex, weighty actions are similarly encoded. A choice between left and right carries no importance to an agent, whereas the choices that interest agency theorists and determinists alike are those that carry the most significance, such as choices for which agents may be held morally responsible.
It is true that the structure of the Soon test does not allow for measuring non-choices (i.e. choosing not to press either button) any more than it allows for complex, weighty decisions. Despite this, there is a sense in which a correlation can be drawn between planning and ‘weighty’ or ‘moral decisions, and between spontaneity and ‘meaningless’ choices. Indeed, Libet found there was a measurable difference in RP between those subjects who planned or considered their actions versus those who acted whenever they felt the desire. In subjects who allowed time for ‘planning’ before an action, the RP occurred earlier and increased at a slower rate than in subjects who acted spontaneously, when RP occurred much later and the rate of increase was much sharper.[14] This is confirmed by a later study, published in 2006, which found late and early components of the RP. The ‘early RP’ seemed to increase in instances where “the agent attends to their intention, is learning a new task, or uses greater physical force.”[15] In contrast to this is ‘late RP’ which increased when actions required precision or strict, isolated muscle use, leading to the conclusion that ‘late RP’ was involved with fine tuning action.[16] Because there is a marked and measurable difference between planned versus spontaneous action, it is not unreasonable to draw a parallel between planned action and ‘weighty’ action. This is because it must surely be the case that planning involves the weighing and measuring of the different options open to a person. So, the content of the action itself would seem to be far less important than whether or not any planning occurred. Because antecedent brain activity was detected in the cases of those who planned as well as those who did not, it seems warranted to conclude that more complex or ‘weighty’ decisions that require more planning would still possess antecedent determined brain activity.
The second potential criticism is that the margin of error could be greater than anticipated because the study relies upon the person’s subjective memory of when they felt they became conscious of beginning to decide. The person could be mistaken, or the apparatus for indicating when the experience of volition occurred could be unrefined.
While this is a possibility, especially with Libet’s older experiments, the most cutting edge and recent experiments would seem to negate such a response. This is because even if it is granted that there is an additional 2 seconds in the margin of error to account for the subject not being sure when s/he decided to choose, there is still a wide enough interval to demonstrate that brain activity occurred prior to the subject becoming conscious of deciding, and that this brain activity can predict fairly accurately the exact choice that will be made. Also, we can postulate that because the Soon test improves upon the Libet experiment and also confirms its findings, similar future studies would also confirm these findings and improve upon the margin of error and possibly the interval, predictability, etc. Equally, because the experiments are testing the timing of volition and action, and we currently do not have a way to observe volition outside of deciding when and where RP factors into volition and consciousness, there is a sense in which these experiments are necessarily subjective.
While the Soon and Libet studies are surely not a knockdown argument against agent causal theorists, I believe they present a significant hurdle to overcome if one is to propose that brain activity does not determine or predict the actions of an agent. They present an equally high hurdle for agent causal theorists who wish to claim that consciousness can inform action without being determined by antecedent physical properties, i.e. neural properties. While not complete and certainly improvable as technology and science develop, they offer demonstrable evidence that at least in all observed instances, antecedent brain activity determines conscious choice. No such evidence seems to exist which demonstrates consciousness arising without any antecedent or encoded brain activity.
Footnotes
[1] O’Connor, “Agent Causation,” 265.
[2] O’Connor, “Agent Causation,” 264.
[3] Details of the study taken from John M Ostrowick, “The Timing Experiments of Libet and Grey Walter,” South African Journal of Philosophy 26, no. 3 (2007): 272, as well as from Sean A. Spence, The Actor’s Brain: Exploring the Cognitive Neuroscience of Free Will (Oxford: Oxford University Press, 2009), 127-140.
[4] Ostrowick, 273.
[5] Ostrowick, 273, quoting Libet, B. 1985. “Unconscious cerebral initiative and the role of conscious will in voluntary action”. Behavioural and Brain Sciences, 8:4 and Libet, B. 1985. “Theory and evidence relating cerebral processes to conscious will”. Behavioural and Brain Sciences, 8:4
[6] Ostrowick, 273-274. As Ostrowick points out, Libet’s assertion that vetoes do not possess antecedent unconscious brain activity seems to imply that they occur out of nowhere- hardly a better option. Ostrowick believes that Libet must either accept that vetoes have antecedent unconscious brain activity, or that dualism is the case.
[7] Sean A. Spence, The Actor’s Brain: Exploring the Cognitive Neuroscience of Free Will (Oxford: Oxford University Press, 2009), 126. Spence also discusses the discovery of a “neural correlate” to Libet’s “free won’t” view of vetoing as a regulatory action, 147-149.
[8] Though not treated in this post, O’Connor’s position can be found in the second half of “Agent Causation” and amounts to the claim that reasons can inform an agent’s decisions but not cause them. For a response to this aspect of O’Connor’s theory, see:
Feldman, Richard, and Andrei A. Buckareff. “Reasons Explanations and Pure Agency.” Philosophical Studies: An International Journal for Philosophy in the Analytic Tradition 112, no. 2 (January 2003): 135-145.
[9] Chun Siong Soon et al., “Unconscious Determinants of Free Decisions in the Human Brain,” Nature Neuroscience 11, no. 5 (May 2008): 543.
[10] Figure 1 from Soon, “Unconscious Determinants,” 543.
[11] Soon, “Unconscious Determinants,” 245.
[12] Soon, “Unconscious Determinants,” Supplementary Figure 6.
[13] Ibid.
[14] Spence, 128.
[15] Ibid.
[16] Spence, 128-129.
Jared,
“This is because it must surely be the case that planning involves the weighing and measuring of the different options open to a person.”
Why should we always think this is the case? If the decision is between, say, wiggling my right finger and wiggling my left finger there may be no factors relevant to the choice of one finger or the other besides my desire to wiggle one of those fingers. The problem with this, as the proponent of the first objection you mention seems to imply, is that desires are passive states. We do not choose our current desires and urges, they come to us, unasked for. So, if the subjects in Libet’s (or whoever’s) experiments were asked to act on passive desires we should expect to find that the subjects did not act in accordance with consciously chosen courses of action. Properly speaking the subjects you cite were not *planning* anything, but simply delaying acting upon their desires.
Andrew,
I am not sure that I follow your comparison of the first potential criticism I respond to and the passive nature of desires.
That being said, the bulk of your response appears to be in line with the first criticism that I outline. Your argument, as it goes, is that there are essentially no factors to be planned for in a choice between right and left – it is merely a capricious choice. However, I do think you are discounting the distinction between a planned versus spontaneous choice. I would think any agency theorists would be alright with the view that a planned choice in the above model essentially amounts to the agent choosing ex nihilo out of the two options out of mere preference or perfect will or some notion like that. Regardless, it seems obvious to me that there is a sense in which I can sit in the experiment and merely react immediately to the urges that arise between right or left, or I can sit in the experiment and say “You know, I am going to decide to click the left button.” This is, as I point out, demonstrated by a difference in RP rates between individuals and maps completely onto whether the subject claimed to allow time for planning or claimed to act spontaneously.
I may be missing an element of your critique or misunderstanding the primary thrust of your point. Please let me know if I have addressed it.
Jared,
Well my point is that if the two choices are essentially indistinguishable, then it’s not surprising that the choice between them is predetermined. This is because the choice will be the result of the desires present in the subject. So, I might have a sudden desire to wiggle my left finger. If I have no reasons to weigh between making one choice or the other then I almost certainly will act upon the desires I have at the moment (to wiggle my left finger). When a subject is told to think about his choice he isn’t actually given any new information about either of the choices, so he doesn’t have new reasons to weigh in making his decision. He’s left with the desires he had before (and perhaps with news desires he finds himself having as he tries to make up his mind about which finger to wiggle). So it’s no surprise that the finger the subject wiggles is predetermined. Desires are not something we can choose, even given libertarian free will. I cannot immediately choose to have a desire to eat nachos. At most I can contemplate the taste, texture, smell (whatever) of nachos and *hope* that this results in the creation of a desire for me to eat them.
Extremely interesting post, Jared. Thank you for sharing.
I do think Brenner has a point. I think also that Brenner’s point explains why vetos do not have the same sort of antecedent brain activity as the “meaningless choice” tasks.
I have got to think about this some more before I comment further.
Andrew,
Thank you for clarifying, I think I understand your point more now. It is true that no substantive information is present upon which the subject can deliberate in examples such as this. However, the answer that such choices are merely desires or urges manifesting in action and not the result of an agent’s choice does not seem to be a particularly strong argument against my conclusion. This is because, firstly, there is a clear difference between those subjects who admitted to planning and deliberating and those subjects who did not. This directly correlates with a slight difference in the RP rate. Further, your argument appears to be that an agent cannot have any reason for choosing right or left simply because there is no information associated with these choices. Again, though, I find it entirely plausible that an agent, upon deliberation, might choose to press the left button based on a memory regarding a childhood school teacher encouraging him to write with his left hand regardless of the hand others were writing with. Would such an example be merely the manifestation of desires outside the agent’s control? It would not seem so.
Additionally, if such simple choices are encoded and thus determined, I do not see a coherent argument for positing that more robust choices would completely upend the findings of these types of studies. On what grounds is this a cogent claim if not the validation of deeply held feelings regarding the nature of personhood and freedom?
Finally, as Kevin rightly points out, vetos are important in this instance. He is right saying that vetos demonstrate a way in which the agent can regulate her behavior. However, as noted in my original post, there are two problems attendant to vetos. The first is that there is no evidence that they too do not have antecedent brain activity which similarly encodes the choice to veto or not. The second issue is that this only represents a negative power, and to base the whole of libertarian agency only upon the ability to stop myself from doing something rather than to will myself to do something is surely (1) strange and (2) ridiculous, since I could then argue that I must constantly be stopping myself from enacting a plethora of dangerous or psychotic actions simply to pursue a more mundane course of action. Similarly, could not any violent or aberrant behavior be justified in that the urges or desires underlying it were too powerful to resists, thereby negating the sort of Moral Responsibility required by the ability to do otherwise?
By way of clarification (Kevin), it is not that vetos do not also have antecedent brain activity – Libet did not test specifically for antecedent brain activity in the instance of vetos, and the Soon study does not appear to include examples of vetos. So, while the studies do not demonstrate that vetos also have antecedent encoded brain activity, the burden of proof would seem to be upon those who argue that vetos (or complex choices) are the “special cases.” This is because I do not see how the mere complexity or desire to veto should not also be in line with what the tests demonstrate about choice and brain activity.
[...] edited and expanded since the original posting. Additionally, a version of this post can be found here at [...]
Congratulations for being featured on Philosopher’s Carnival, Jared!
http://usuphilosophy.com/2011/04/25/welcome-to-philosophers-carnival-124/
Thank you for the interesting discussion on the recent results, which I found enjoyable even as far as from Italy. Unfortunately, technically speaking these do merely show that there are necessary computational antecedents of decision-making which occur in brain modules performing unconscious processing of data. The whole issue of the switch in predicitive capabilities between different brain regions and several other considerations you will allow me to not tackle here, leave room for plenty of other interpretations. Regards from afar,
F.C.
Hi F.C.
Thanks for reading!
As mentioned in the study itself and my original post, that prediction is even possible based on brain mapping is due to such antecedent brain activity encoding the content of the choice. As such, this is not mere preparatory processing of data in anticipation of the agent making a choice free of any causal efficiency on the part of antecedent brain activity. Rather, it appears there is more going on behind the scenes for a notable period of time before a choice enters the awareness of the test subject than we might initially think.
As I concede above, there are many ways in which these types of studies can be improved. And such studies are being improved as we speak – see here (http://agencyandresponsibility.typepad.com/files/neuroscience-and-free-will-nature-aug-2011.pdf) for an article on this very controversy, but also for details on some of the upcoming studies that seek to address many of the lacunae in these studies thus far.
That being said, that the predictive capacity appears to shift brain regions but maintains a similar percent of predictability would seem to reinforce my conclusion in the original post. For, if we have a similar predictive capacity in the frontopolar and precuneus regions prior to consciousness of a decision as that of the motor cortices after consciousness, then in what sense is consciousness of a decision necessary for predicting the choice? No doubt as we learn more about the functions of these particular areas of the brain, we will be able to extrapolate more accurately from studies such as these and those to come.
[...] Libet Revisited: Reduction and Prediction in Neuroscience (UNFSPB Guest Post) [...]