Blog Coordinator

« The Main Argument, continued | Main | I'm Back: Revisionism vs. Classical Compatibilism, part I »



Feed You can follow this conversation by subscribing to the comment feed for this post.

Hi Peter, Great post! I also enjoyed chatting with you about all this stuff last week in Tallahassee.

Just a thought about something you say under (3).

As you know, I like your idea that in order to conclude that A causes B, reparameterizations of B might need to be taken into account. However, you say: "On this view, standard interventionist and Newtonian models of causation are a special case where B places no conditions on input from A." If by standard Newtonian models you mean something like the physical connection theories of Dowe and others (i.e., theories that emphasize transfer of energy, etc.), then maybe you're right. But I'm not sure what you say is inconsistent with interventionism. All that interventionism requires, roughly speaking, is that that there's some value A can take in a model such that for at least some state of the model there's an intervention on A that would result in a change in the value of B. And that might partly depend on a reparameterization of B.

But I think you're right that less roughly speaking, in terms of the details of a specific interventionist view like Woodward's, this isn't made explicit.

In philosophical terms, here's one way of stating the sort of requirement that I take it you want for your "criterial causation."

When we ask whether A is an actual cause of B, Woodward says we must first "screen off" the causal influence of other variables — like C — that feed into B by holding them fixed at their actual values. Then, if an intervention on the actual value of A results in a difference in the actual value of B, we conclude that A causes B. Christopher Hitchcock has a nice way of putting this. He says that in normal counterfactual (difference-making) accounts of causation, we ask whether the following counterfactual is true: "If A had not occurred, then B would not occurred." Here, the antecedent of the counterfactual only makes a stipulation about a single event (A). But in interventionist difference-making accounts, we ask: "If A had not occurred, but C stayed the same as it is, then B would not have occurred." Here, the antecedent of the counterfactual makes a stipulation about *more* than one event (A and C). I take it you're saying we need to include a third conjunct in the antecedent, namely one that says something about how B is parameterized and whether it is parameterized in such a way as to be receptive to patterns in A.

I think that's a legitimate amendment that might need to be made to Woodward-style interventionist views.

Oh, and briefly, I agree that prospection and imagination in deliberation are exactly where the action occurs when it comes to why people *believe* in free will (particularly libertarian free will — for those who believe in it).

Hi Oisin,yes thanks, that is put very well. It is just that small amendment that would, I think, make Woodward-style interventionist accounts of causation more complete and even better. It seems to me that Woodward is pretty much describing in an idealized form what scientists usually do as part of their standard operating procedure, which is to try to hold all variables constant, then varying one 'independent' or 'manipulated varible' to see what effects there might be on one or more 'dependent' or 'response variables.'But changing the code or parameters or criteria that B uses to decode, interpret or respond to input is a manipulation that might make no apparent changes to A or any other variable in the system for long and uncertain spans of time, until just the right pattern comes along. For example, the Mossad might program a cellphone to explode only when a particular number, known 'only' to their target, is dialed. It might take years to get this phone into the hands of their targeted Hamas leader. But when he dials it, he dies. This kind of reparameterization of B need not have immediate noticeable or measurable effects within the system. But it is a causal intervention nonetheless.


I enjoyed reading your overview. Regarding section (3), would it be appropriate to model B as a thought and A as a neuron, and the thought is what controls the synaptic weighting of the neuron?

Hi Peter,

Many thanks for your précis. A couple of things:

1. I'm wondering how the indeterminism in your model, what you say is necessary for "strong" free will, makes the agent more responsible than in a deterministic account. In your model, things could have turned out otherwise in the actual situation (as opposed to counterfactual) because of quantum random influences on neural firing. But it seems to me this breaking of the deterministic causal chain doesn't help to make the actual choice a function of the agent's characteristics, what we normally look for in assigning origination for action and thus responsibility.

2. You say “Qualia are necessary for volitional mental causation because they are the only informational format available to volitional attentional operations.” Presumably the informational format of neurally instantiated volitional attentional operations associated with qualia can (and someday will) be specified in neuro-computational terms, in which case we needn’t appeal to the subjective qualitative format of qualia as playing a causal role. Of course if these two formats turn out to be fundamentally the same thing in some sense (that case hasn’t been made yet I don’t think), then the causal role of qualia in your avowedly physicalist paradigm will have been secured. But at the moment, in 3rd person accounts of behavior it’s the neuro-computational format that does the explanatory work, so mental (phenomenal) causation is yet to be established, seems to me.

And a followup on Tom's queries: in the spirit of Manuel's tracing problem, exactly (or near enough) where/when in the neurology and/or qualia are we to find the location of buck-stopping agency?

James, in my book I take B to be a post-synaptic neuron and A to be a pre-synaptic neuron. But that is just an instance where the criteria realized in B (in this case in its synaptic weights) dictate whether A will be causal of B’s behavior, and if so how (in this case of B’s firing or not, and if so, in which way it will fire). But in the phone example I gave in response to Oisin above, A would be the Hamas leader dialing various phone numbers, and the telephone B’s response would depend on that input (if the secret number, it explodes, otherwise it acts like a regular cellphone).

Of abiding interest and puzzlement to Psychologists and Neuroscientists is exactly how a thought, say that I need to recall any high school classmate’s name, can trigger subsequent neuronal events such as recall of one of the many possible classmates who could have been recalled given the criterion that it must be a classmate. I predict, but can’t prove for lack of adequate technology, that the neural realization T of such a thought, presumably realized in a prefrontal working memory area, can parameterize (change synaptic weights on) neurons B, which will then assess elicited inputs from memory areas, such that B then sends a classmate’s name back up to the neurons that realized T. This is experienced as a name just popping into your head. I can elicit this experience in you by asking you to think of a mammal that lives in water. A fraction of a second later an image or word or proposition just pops into your consciousness. It might have been a dolphin or an otter or some other aquatic mammal. Had we ‘rewound’ the universe to precisely the moment when B was parameterized to be satisfied only by inputs that meet this criterion, a different aquatic mammal might have come to mind. But it specifically met that criterion. No one deeply understands the details of how that process works at a neural or synaptic level, but it is remarkably fast and accurate. I take some educated guesses about how it might work in the book.

Tom, this parameterization of B also gets at why I think that “breaking of the deterministic causal chain” DOES “help to make the actual choice a function of the agent's characteristics.” It was the executive circuitry that reparameterized B such that only some aquatic mammal would suffice. So the outcome is not determined (it could have turned out otherwise) and it is not utterly random (it had to be an aquatic mammal). There is therefore a middle path “betwixt chance and an absolute necessity” contra Hume cited above. To the extent that one equates such executive circuit behavior and decision-making with an agent, then the agent is responsible that it was some aquatic mammal that was recalled rather than all the other conceivable things that could have been recalled if recollection had been truly random. Of course, one could then argue that the executive circuitry had to come to the conclusion that a classmate’s name was needed. But microscopic noise can be amplified to randomness in spike timing for executive circuits as well. Rewind the universe to the moment before T, and you might instead end up with the thought ‘I need to recall the name of some past lover.’ Ultimately executive decisions will be driven by some factors beyond the control of those executive circuits, such as innate wants and desires, including things that count as smelling good or bad, and learned wants and desires, such as what counts as prestigious or as an in-group vs. out-group member. So I am certainly not saying that we are utterly free to want what we want. I am just saying that given what we want, the conditions for having a strong free will are met by this kind of neural coding scheme that emphasizes reparameterizations of B. (Tom, I will tackle your qualia question after dinner :)


Thanks for your reply.

If post-synaptic neuron B is able to influence the synaptic weighting in pre-synaptic neuron A, and if that were to happen in a coordinated and time-phased manner to *billions* of neurons B and A during a thought process, might it be reasonable to believe that the emergent “thought” effectively changes the synaptic weighting of a large distributed set of neurons in a coordinated manner? Here’s where I’m going with this…

It sounds to me like human thoughts are caused by billions of neurons firing in a coordinated manner, while at the same time the individual neurons don’t control how the emergent thoughts change synaptic weightings of the neurons. In other words, a thought is a new higher-level entity that exerts control, and that control isn’t predetermined. Could that be a reasonable assessment of where our free will originates from?

I believe that it’s possible for mental events to be causal within a physicalist paradigm (as you suggest). Doesn’t that imply human thoughts exert some kind of physical force? Based upon your overview of your book, I would think that it’s reasonable to believe that thoughts exert new emergent forces – not forces that simply result solely from a direct sum of the four fundamental forces of physics. Yes, the *net* affect inside a physical brain abides by the laws of physics, but a thought may produce new emergent forces that add into the mix in real time, thereby summing with other preexisting forces and affecting the net sum of forces, thereby affecting the path of reality, and thereby producing free will in the “strong sense”, as you’ve referred to it.

If the above is reasonable, isn’t it also reasonable to believe that some kind of “transcendence” occurs when a thought exerts forces in a downward-causation manner and thereby affects the synaptic weighting of neurons? In other words, if a thought exists at the wave level of neurological activity, then wouldn’t the forces which emerge at the wave level need to transcend downward into lower-level fields in order to affect the neurons?

I’m wondering if the phenomena you’ve discovered illustrates that life is fundamentally indeterminate.

In response to Tom’s question about the causal role of qualia, let me flesh out the view of qualia I develop in my book a bit. I argue that the link between qualia and volition (via volitional attention) is very close: Qualia are those informational structures that can potentially be volitionally attentionally bound and tracked over time, and everything that can be so tracked if need be includes all current qualia (i.e., all that is now experienced). On this account, all that is presently experienced includes that which is attended now and that which could be volitionally attended if we shifted our attention to it in the next moment. If only conscious contents can be volitionally attentionally tracked, then consciousness or experience is, conversely, the domain of representations that permit the possibility of volitional attentional tracking and manipulation. On this account, unconscious operations may be driven by input rather transiently, on the timescale of the rapid allocation and rapid decay of exogenous attention. In contrast, conscious operations may be driven by input in a sustained fashion, on the timescale of the slow allocation, but slow decay of volitional attention.

What is the informational format of qualia? The outputs of automatic, stimulus-driven, early modular processing comprise a precompiled, preinterpreted, and preevaluated account of events and objects in the world—as well as of the states and needs of the body. These mid-level outputs are made available to brain areas that can make cognitive inferences and plan to do something given these facts, whether motorically or internally by, for example, volitionally shifting attention or exerting effort to recall a name. Qualia are those representations that can be (or now are) operated on by volitional attention, giving rise to the possibility of volitional attentional tracking, which, I argue, cannot happen in the absence of consciousness. Because certain operations can take place only over conscious operands, and motor acts can follow and enact the conclusions of such operations, such mental operations can play a necessary causal role in subsequent thoughts and motor acts.

Why do we experience anything at all? Why not just have a vast array of reflexive zombie agents for every possible input–output contingency? For a complex animal, there would be countless such contingencies. A brain could not store infinitely many reflexive input–output mappings, even if it could know them all in advance. In addition, reflexes, while fast, are not flexible in their mapping of input to output, are locally and slavishly driven by the input stimulus, and do not permit dynamic, open-loop selection among alternative options. In contrast, experience is closely linked to both working memory and attentional processes that do permit dynamic, open-loop selection among alternative options.

Another part of the answer is that a common, volitionally attendable format is required so that the outputs of different subsystems can be evaluated by a common set of criteria assessing input, which, if satisfied, releases a decision to act. Qualitative, subjective experience is the common output format of numerous relatively modular subsystems that process fundamentally different types
of input. Having a highest level of assessment of multiple lower-system outputs in a common format permits the overall system to find a solution that maximizes global benefit to the animal, as opposed to locally, within the representational space of any submodule. A common format is required so that all submodules’ outputs can contextualize one another, can be “time-stamped” as occurring in the same moment, and can be stored in the quasi-experiential, spatiotemporal format of episodic memory. Having experience in a common qualia format allows executive, planning, attentional, and working memory processes to have all relevant information about goings-on in the world and body at one time. In the absence of a common format, the relevance or salience of an output from one module, say, redness, would not be comparable to the output of another, say, hunger. Their relative salience, importance, or priority would not be rapidly decidable. This shared format allows executive processes to consider, compare, track, and select from among possible courses of action in a unified way.

Why are qualia always in a first-person format? Because they are only accessible to and operatable upon by a single volitional attentional operator, namely the one in the experiencer’s brain. Moreover, they typically provide planning circuits information about the experiencer’s body with a singular perspective on events provided by the senses. But if I could link our brains by an artificial corpus callosum, or better yet, if I could send appropriate inputs to your neurons from my neurons via photons, you could in principle experience my qualia and vice versa, just as the left and right hemispheres appear to accomplish this via the real corpus callosum.

But if volitional attention can operate only over conscious operands, does this mean that consciousness per se is causal of motor acts or thoughts? No, it does not. It is generally the case that certain operations can take place only within a particular space. To use a metaphor, I can only turn (an operation) a screw (an operand) with an electric drill (the operator) in three-dimensional space. While this space affords me the possibility of carrying out this action, it is not itself a sufficient cause of any particular action. Similarly, consciousness need not itself be sufficiently causal. It can be thought of as a space or domain of a particular type of information. Certain operations, particularly those associated with volitional attentional manipulation and the tracking of representations accessible to this operator, may be possible only within the conscious domain. Primary causation lies with the “operation of the operator on the operand” in the conscious domain, not with consciousness per se. An alternative metaphor is that the electricity needs to be on in order to afford the use of an electric drill to screw in a screw, but that the electricity is on does not guarantee that any operations requiring electricity will be carried out. Consciousness, in the sense of everything experienced now, would, according to this metaphor, play a necessary but not sufficient role in certain kinds of operations. Consciousness is necessary but not sufficient for volitional attentional operations, while the presence of volitional attentional operations is sufficient for establishing consciousness in people if not other species. For example, akinetic mutes might have damage to medial prefrontal areas associated with ‘willpower’ and plan implementation. They continue to have qualia or experience, but sit listlessly, doing and saying nothing, because they have lost volitional operators, while still having the necessary but not sufficient operands for planning and action.

James, to believe that information, mental or otherwise, can be causal, one has to be willing to take the strong stand that (physically realized) information can influence events at the root-most physical level to turn out one way versus another. Present physically realized informational criteria for neuronal firing influence which possible particle paths can become real by only allowing possible physical causal chains that are also informational causal chains to occur. Neurons place physical criteria on physical inputs that realize informational criteria placed on informational inputs. Synaptic weights limit what inputs can trigger firing. But which particular set of spike inputs ̶ and thus what information ̶ will make the neuron fire is unforeseeable, so long as the physical/informational criteria for firing are met. Neuronally and genetically realized information are both causal qua information, because the physical systems in which they are realized are organized to permit only the occurrence of physical causal chains that are also informational causal chains. Those possible physical causal chains open to particles that would not result in an informational causal chain are excluded by physically realized informational criteria that must be met before the release of a physical action. Among neurons, these physical/informational criteria for the release of a postsynaptic spike are placed on the coincident arrival of presynaptic spikes. If neural causal chains are also informational causal chains, and the same informational causal chain is realizable in countless different neural or particle causal chains, then the more parsimonious model is one of information causing information. Yes, there must always be some physical realization of information, but under physical/informational criterial causation, which one it happens to be is irrelevant so long as informational criteria are met. Indeed, chains of successive informational criterial satisfactions and criterial resettings afford the physical realization of both downward mental and downward genetic informational causation. Against eliminativists, if macroscopic settings, such as synaptic weights that realize informational ‘pattern detection’ criteria, bias which subset of lowest level events can be realized from among all possible such lowest-level events, that is top-down physical causation. In fact it is top-down (physically realized) informational causation. But there is no need for thoughts to exert physical forces above and beyond the forces exerted by the physical substrate in which they are realized. Mind and information must always be physically realized, assuming, as I do, physicalism. The key thing is that thoughts are not realized in amounts of energy, but in patterns, and patterns only become causal in a physical system if there are decoders that do something upon detection of patterns to which they have been parameterized to respond. Please see my October debate with Neil, particularly toward the end of the thread for more on how information is top-down causal within a physicalist paradigm.

I believe that in time we will come to see qualia as a type of information, in the specific format required for attentional allocation and manipulation in working memory, and for associated executive decision making processes. I realize most philosophers take qualia to be the irreducible subjective feels or phenomena of experience. For example, Ned Block might say that I am talking about access consciousness but not phenomenal consciousness. Access consciousness is sometimes delimited as comprising those contents that one can report voluntarily or access voluntarily, whereas phenomenal consciousness is what is experienced. On my view, the distinction between the two types of consciousness melts away because they are united by the fact that whatever can is accessed/reported can be volitionally attended, and whatever is volitionally attended is accessible/reportable, at least by humans, and whatever is volitionally attended is experienced subjectively. The first-person aspect of qualia can be accounted for, as I mentioned above, because they are only accessible and manipulable by one’s own volitional attentional operations. Unconscious contents, in contrast, would be those that volitional attention cannot access or manipulate in the mental workspace.

Take everyone's favorite example, pain. I argue that it is information in a particular format (operatable upon by volitional attention) that has particular roles and properties within a dynamic information processing architecture, such as receiving high prioritization, invoking thoughts and behaviors to make it stop, and automatically drawing attention to itself. Let's remove these information processing properties but keep the information associated with the qualitative sensation of tissue damage. Would it still hurt if we no longer incessantly attend to it, prioritize it or try to make it stop? In that case we might say 'I am in agony but it doesn't hurt or bother me.' Wouldn't that be strange? Would we still call such a sensation 'pain'? Certainly it would lose its similarity to other instances of what we call 'pain.' What if I now also remove the information associated with the location of tissue damage, being in, say, my left ankle joint, and also remove information about what kind of tissue damage it is. I would argue that if we remove all the information processing properties and all information associated with what we call pain, there would be nothing left over to call a 'quale.' For those who want to reify qualia as something beyond information, I would ask, what exactly is left over after we remove every informational property or aspect from pain (or any other quale)?

How could we empirically or even conceptually settle the matter between those who believe that p-consciousness is something beyond information accessible to attentional manipulation and those, like me, who believe that the p- vs. a-consciousness distinction collapses if information is thought of in a more sophisticated way than Shannon's quantification of reduction in uncertainty? If no experiment can settle the matter, how should we as a field proceed? If Ned comes back to me and says that a 'subjective feel' or 'qualitative essence' is what is left over after we have removed all information and informational attributes (such as prioritization etc.) from pain, I would ask Ned "What is a subjective experience that carries no information? Please give me an example." Also, what would that be like, subjectively?

The "precompiled" information made available to possible attentional selection is highly preprocessed, and has been converted into post-constancy and rather Gestalty formats like 'this material is solid,' 'this object is this far from me on the groundplane,' 'this surface has this pigment.' Qualia are experienced because they comprise the very (but entirely informational!) 'virtual world' that voluntary attention can operate on.

Hi Peter. Since I doubt I can add much on the neuro-scientific side of things, will you humour me if I just ask a couple of fairly unsubtle questions to try to link it to the philosophy I know better?

Firstly, you've written about why you think we can have free will in an indeterministic world, if the details at the level of neurons are as you say, and it makes a lot of sense; you also finger indeterminism as the key to achieving genuine mental causation. But I suspect that many philosophers will be more interested in asking "Could we get the same thing even if determinism were true?". Say that, over the next couple of decades, neuro-science makes great progress and it turns out you were completely right about how human beings work, and you're sitting in your office feeling great about having shown how we can have mental causation and free will. Then a major discovery in fundamental physics is announced over the radio - it turns out that quantum theory was wrong; a new, better theory has been discovered, which shows how all the predictions of quantum physics about phenomena that could affect our neural processes were almost exactly right, but is nonetheless completely deterministic. Are you actually going to change your mind at this point, and come to think that we never had mental causation and free will, or will you think the fundamental physics makes no difference, as long as you were right about neurons and information processing? (I suppose it's worth pointing out that Everettians think that quantum physics is already deterministic, but maybe you think it provides the right kind of 'alternate possibilities' even so.)

Secondly, on qualia, do you think that there could be a being that felt the sensation I would associate with being tickled when they stubbed a toe or grazed an elbow, and felt the sensation I would associate with pain when they were tickled? It seems to me that tickling and pain produce similar reflexes, are limited to locations, come in degrees, etc. - if there is an informational difference, what do you think it is? Alternatively, do you think there could be a being that felt the pain I would associate with toothache when they tore a muscle, but felt the pain I would associate with a torn muscle when they had toothache? The two pains will be even more difficult to distinguish informationally than pain and tickling. (As far as I can tell, the only difference in information between muscle pain and tooth pain is that one happens to go with muscle damage and not tooth damage and the other is the opposite, so you could swap two kinds of pain in some being and leave them with the same information, but that means that the information a pain state carries doesn't exhaust its phenomenal character.)


I think I’ve identified the fundamental roadblock that I’m running into regarding your statement “there is no need for thoughts to exert physical forces above and beyond the forces exerted by the physical substrate in which they are realized”.

In section (3) of your overview, you mention “the constant reparameterization of B” as being key. If everything that’s located within physical entity B is controlled *solely* by the four fundamental forces of physics (4FFOP), then wouldn’t it be fair to say that the reparameterization of B is predeterministic in nature? I’m thinking that in order for “the reparameterization of B” to be a concept that supports the existence of free will, new emergent forces *must* exist within B thereby allowing the activity within B to be *different* than what would occur if the reparameterization of B was controlled only by the 4FFOP.

I agree with you that biological systems are indeterministic, but I also believe new emergent forces are the essence of life.

Peter (if I may), I thank you for your careful exposition. But I have a question: so much of your exposition is cast in the passive voice of various physiological and conceptual parameters of what may be possible for agency. I don't wish to be a nag, but what exactly (or again near enough) do you take to be the indeterministic ground of agency in specifiable spacetime terms (since you apparently wish to ground agency in a way that relates to empirical references)? I just want to understand your central claim(s) of what constitutes the FW incompatibilist agent as clearly as possible.

Alan: I can't speak for Peter, but I would suggest that the most natural indeteriministic ground for agency in Peter's account, in specifiable space-time terms, is presumably the quantum wave-function and its (apparent) "collapse", the mechanics of which is well-understood outside of brains but not within (see e.g. ).

CJ, thank you for your questions. I feel a parallel lack of comprehensive knowledge, but in philosophy, since my background was in physics/math, before switching into neuroscience over 20 years ago. I have been teaching myself philosophy for the past decade or so, since it became clear to me that the basic ideas of my field, like 'information,' 'mind,' 'causation,' 'representation' and so on were not well-defined, at least as many of us were using them. Also, I became obsessed with the NCC issue and realized we neuroscientists needed to sort things out conceptually, and not just via experiments. But as I kept reading, I could see parallel debates and uncertainties in your field. (I really have to thank Thomas for having a blog where an outsider like me can interact with philosophers like you. The more of this kind of cross-field fertilization/debate the better. And please correct me if I commit any gross or minor philosophical errors! I want to learn).

In reply to "Could we get the same thing even if determinism were true?" I emphasize in my book that the neural code could be one that accomplishes information processing in part via rapid synaptic weight changes even if the world were deterministic. It is just that the weight of evidence, reviewed in my appendix one, supports the conclusion that the microscopic domain is indeterministic. Then the question is twofold. First, can, and if so, how can microscopic randomness be amplified to a macroscopic domain, like spike timing? Second, can some neural coding scheme place constraints on randomness, such that it is 'harnessed' for information-processing goals to get around the danger of randomness alone being 'unwilled.' I go into details that answer both questions in the positive.

Even though I think the quantum domain is indeterministic, and that this indeterminism is in fact amplified to a macroscopic domain of neural behavior, a weak version of free will is indeed compatible with determinism. Since compatibilism is just a consistency claim, as Oisin said to me last week in Tallahassee, one can be a compatibilist even if one thinks determinism is not the case. Compatibilist free will is weak because events could not have turned out otherwise (ruling out such extravagances as changing the laws of the universe or changing the past). But under a strong version of free will, things have to be and have been able to really turn out otherwise, and for that form of free will, indeterminism is necessary, but not alone sufficient.

Now I have a question for you and other philosophers. Is the issue of determinism vs. indeterminism empirically or conceptually even settle-able? If you take any subset of the universe, outside influences, such as cosmic rays, will incessantly impinge on and alter this non-closed subsystem. Therefore events within the subsystem are never sufficient to account for what will happen within the subsystem subsequently. Even if the universe as a whole is deterministic, for any part of the universe, such as the brain, this kind of 'as if' indeterminism will reign. But to establish that the whole universe is deterministic would require measuring all particles in the universe at once. But since we are in the universe, there is a Goedelian problem, let alone a Heisenbergian problem. So we cannot settle whether the universe as a whole is deterministic or not. If we have an 'as if' strong free will or a 'real' strong free will may hinge on whether synaptic noise that gets amplified to spike timing randomness is determined or not. But if something cannot be settled empirically or conceptually, such as whether or not there are multiple universes, branching a la Everett or not, have we not hit the limits of reason and knowledge? Why lose sleep over something that cannot be settled in principle? Or do you think the matter can be settled empirically or conceptually?

CJ, concerning inverting qualia or switching them, I would think that if there is a difference in qualia, there must be some difference in information or informational properties within an information processing architecture. For example, I cannot tickle myself, because tickling concerns touch to vulnerable areas that I do not cause myself, whereas I can cause myself pain. I suppose if neuronal inputs were switched, one could come to perceive tooth damage as pain in one's toe. There are actually some remarkable scientific and 'natural' experiments along these lines.

(1) Mriganka Sur and colleagues at MIT carried out 'rewiring' experiments, reviewed here (, where retinal inputs were sent to auditory cortex. Even after visual cortex was ablated, the modified ferrets could make visual discriminations using auditory cortex! Now, whether these ferrets had visual experiences or not is perhaps not knowable, but they certainly had visual information realized in neuronal activity that normally would process sound. If we could do such experiments in humans, they could just tell us whether associated experiences were visual, auditory or otherwise.

(2) An example of 'cross-wiring' in humans happens with phantom limb patients. After losing an arm, cortical neurons in somatosensory areas that previously received hand touch input no longer do. Adjacent face touch processing areas expand their cortical turf to include neurons that previously took hand input. Such rewiring leads to changes in qualia, because a touch to the face can be experienced as a touch to the non-existent hand, as shown by V. S. Ramachandran.

(3) A final example is trichromacy. Most mammals have two cones, so might have color experiences rather like that of a color blind man who lacks one cone type. Mollon and others have hypothesized that trichromacy in fruit-eating primates may have evolved because redness had already co-evolved between trees and birds as a signal for fruit ripeness. When the third cone type arose as a mutation of the second cone type, presumably no new cortical machinery was necessary. The informational input went from two to three dimensional, and so, presumably did the experienced colors. Even more interesting, Mollon and others have studied a woman who has four cones and exhibits behavior consistent with a four-dimensional color space. (

I believe such findings are consistent with the view that qualia are informational structures with very particular properties, but entirely informational nonetheless.

James, I do not see what is to be gained in positing new forces or new laws of physics if the existing laws of physics can be shown to be adequate to account for the central-most issues of mental causation, free will and consciousness. I worry that positing new forces is like positing elan vital. Once we came to understand the informational code underlying genetic inheritance, most people came to see life as informational and dispensed with elan vital. I think if we come to understand the informational code underlying mental/neural processing and causation, we will dispense with notions of dualism or mind-forces. Of course, physicalists like me might be wrong, but in the spirit of Quine, I think we should have as spare and desert-like a metaphysics as we can muster, at least until we reach a point where we have no choice but to accept the inadequacy of our assumptions (e.g. monism, supervenience, adequacy of current physics for modeling the brain, skepticism toward claims of a central role for strange quantum effects etc). I am the first to admit that the universe is not well understood by humans, and that the universe is likely to be much stranger than we (can) conceive of it. At the very least, it has to be strange enough that human consciousness could exist. But, for now I will cast my lot with monism and the known forces and (indeterministic) laws of physics. Physics is plenty strange enough already. While I do not believe that nonlocality, quantum coherence, or tunneling play any role in neural information processing (the brain is just too warm!), I do believe that atomic level indeterminacy, which is a quantum domain effect, is enough to bring some of the strangeness of the quantum domain to the domain of human thoughts and creativity.

In defense of "other philosophers" and settling things conceptually.

Either the determinism/indeterminism issue (parsed as absolute or relative to systems) is relevant to free will/responsibility or it is not. If it is then incompatibilism is correct; if not then compatibilism is correct, or skepticism is correct if the the whole issue about FW/MR is somehow itself wrong.

There are three thus logically possible positions. Just to be clear. Each position is thus charged to make its position as clear as possible.

And thank you Marcus for the reference. But the nature of wave-collapse or associated settlement of quantum-involved circumstances by some agency seems to me to be the issue at hand in this thread. I just want to know as specifically as possible what that proposed agency is, and not in terms of mere possibilia.

Alan, some philosophers whom I have read want to locate agency in a person. But what counts as a person? If I remove arms and legs, the person clearly remains the same person. Same goes for most parts of the body, as long as a function is replaced (e.g. of kidneys by a dialysis machine). But alter the central nervous system and the mind can be changed, and with it its agentic attributes. So I would place agency in the central nervous system; not just one agent, but several agentic or executive circuits (meaning they can make decisions that can govern behavior or thoughts, and can initiate action). These agentic circuits operate in parallel while mutually interacting. I would not agree to Freud's notion of id, ego and superego, nor would I agree with the ancient metaphor of a reason-like rider trying to govern emotion- and desire-like horses, found in Plato or the epic of Gilgamesh. But these old metaphors were onto something. Different circuits process different inputs and appear to push behavior to different sorts of ends. Some of these afford conscious volitional control, while others operate unconsciously. In modern neuroscience terms, there are specialized circuits for particular forms of behavior/thought. I suspect no one would want to identify their agentic self with unconscious executive decision-making processes. Instead most of us identify our agentic self with that which has access to and can manipulate the contents of consciousness by, for example, shifting the 'spotlight' of voluntary attention. One of the key agentic circuits, but by no means the only one, is the fronto-parietal circuitry that underlies voluntary attentional shifting, allocation, and manipulation of operands within the mental workspace. Indeed, I would say that one of the free-est acts of a free will is the capacity to shift voluntary attention to any representation and maintain attention there despite other possible information processing streams vying for, well, attention.

Marcus, James, I do not think agency or consciousness collapse the wave function. Here are two relevant paragraphs from my book:

A veritable cottage industry has emerged among physicists who
have suggested that mental events somehow follow from quantum domain entanglement and nonlocality (e.g., Hameroff, 2001; Penrose, 1989, 1994) or electron tunneling (Walker, 2001; cf. Macgregor, 2006). Such claims are improbable (Tegmark, 2000; Grush & Churchland, 1995: but see Penrose
& Hameroff, 1995, for a rebuttal) and are not needed to account for mental causation. In contrast, criterial detectors, such as receptors or neurons, can operate in the domain of ordinary temperatures where the kind of coherence that would be necessary to realize entanglement would be made incoherent (Koch & Hepp, 2006). There is no need to invoke quantum nonlocality, superposition, entanglement, coherence, tunneling, quantum gravity, or any new forces to understand informational causal chains in the brain. Criteria can be realized in the input–output mechanisms of relatively large scale, high-temperature entities, such as receptors or neurons, in the absence of nonlocality effects. What is needed, however, is some degree of noise in the system that arises from amplified microscopic fluctuations that manifest themselves as randomness concerning the timing of EPSPs and IPSPs and therefore neural dynamics. Because of such noise at the synapse and within neurons themselves, there is no guarantee that identical presynaptic input will lead to identical postsynaptic output, even if time could be “rewound” and initial conditions were truly identical. But noise could also be introduced by external factors, such as, say, noise in perceptual inputs, or cellular damage due to free radicals or cosmic rays, or many other possible causes that have nothing to do with nonlocal quantum level effects. While I argue that noise can be harnessed for the purposes of generating novel solutions using criterial causality, this is a far cry from notions that nonlocal quantum-level effects are in some mysterious way responsible for mental events. It is improbable that any of the strange, nonlocal quantum coherence effects can have any influence on how neurons behave, or how consciousness or information is realized in neural events. The brain is, simply put, too “warm” to support this kind of quantum-domain coherence, and synapses are too wide to support electron tunneling. Just because some quantum effects are mysterious and
the physical realization of mental phenomena is also mysterious does not mean they are related. In short, I doubt that quantum-domain effects— beyond the variability in neural dynamics introduced by amplification of microscopic fluctuations—are required to account for how information is processed by neurons.

Some physicists and philosophers have argued that quantum indeterminism permits a gap in physical causal chains that can be exploited by consciousness to bias which possibilities become real (cf., e.g., Hodgson, 1993; Penrose, 1989, 1994; Stapp, 2004). The view developed here is unlike such views because consciousness, in the sense of experience, is not seen to play a necessary role in determining which possibility is actualized. Rather, consciousness, and the entertainment of possible scenarios and courses of action in working memory, plays a role in changing the criteria for firing on neurons that might lead to future mental events. In other words, experience and online manipulation of representations in working memory allow the potentiation of future mental events and actions, not present ones.

Peter, thank you so much for your last comments--they are very helpful in seeing where you;re coming from. I need to digest them a bit.

The comments to this entry are closed.

Books about Agency

3QD Prize 2014: Marcus Arvan