Russell, Newman, and the mind-body problem

Bertrand Russell's The Analysis of Matter has had a resurgence of interest in recent years, both among philosophers of science interested in structural realism, and among philosophers of mind interested in the mind-body problem.  Russell's basic idea that science characterizes the physical world in terms of its structure but not in terms of its intrinsic nature has seemed to some philosophers (Russell himself, and more recently Grover Maxwell, Michael Lockwood, and me, among others) to leave the door open to a close tie between the unknown intrinsic properties at the basis of physics and consciousness itself.  This leads to the view that I have called Russellian monism (or type-F monism, in "Consciousness and the Place in Nature").

At the same time, Russell's structuralism suffers from a well-known difficulty, pointed out in 1928 by the mathematician M.H.A. Newman in "Mr. Russell's 'Causal Theory of Perception'" (JSTOR link), and revived in a 1985 article by Demopoulos and Friedman.  A purely structural description of the world, saying that there exists a relation under which entities in the world are related with a certain structure, can be satisfied by any set of the appropriate cardinality, by defining an appropriate relation on that set.   So this sort of purely structural description seems near-vacuous.  Some philosophers take this to be a devastating problem for structuralism (e.g. Michael Friedman raises it as a deathknell for Carnap's Aufbau), while others (e.g. Putnam in his model-theoretic argument) use it to raise problems about realism and reference.

I received an interesting e-mail from Moises Macias, an undergraduate at the National Autonomous University of Mexico, asking how one can reconcile this problem for structuralism with its popularity in the philosophy of mind.  This led to some useful correspondence.  Since this is an issue that quite often comes up in discussion, and it's one of my favorite issues, I thought I'd post some of the correspondence here (after the fold, with text from Moises indented).

Moises:

How come if Newman's criticism is so devastating then so many arguments and trends in contemporary philosophy of mind (Lockwood, Chalmers, Maxwell) start off from Russell's argument (that physics tells us only about the structure of the world, the dispositional properties of its objects and not of its internal properties) to the conclusion that perhaps Russell was right in claiming that we have knowledge of the intrinsic properties of at least one physical system, namely the brain?  Lockwood, Maxwell, Stoljar, Nagel, Galen Strawson, Unger and you all present pretty good arguments for at least considering seriously the Russellian position, a position which, if I understand correctly, could only work if Newman's argument is not as destructive as many other philosophers claim it to be.  Also, if memory serves me right, van Fraassen has an interesting and somewhat sympathetic discussion of structural realism in its own right which can be found in his web page. How come no one, as far as I know, has examined Newman's arguments in the discussions concerning Russellian theories of the mind, and vice versa?

Good question.  I'm sympathetic with Newman's argument, but I don't think it applies to the sort of view I advocate.  Newman's argument applies against "pure structuralism", where every term in the theory is in effect defined structurally, and where there are no primitives that get their meaning independently.  The view that Lockwood, Maxwell, and I advocate is not nearly that strong.  The claim is that terms like "mass", "charge", etc, are defined structurally.  But there may be other non-structurally-defined terms.  E.g. "cause", "law", "experience", and perhaps "space" and "time" (though I'm doubtful about those two).  Some terms like these will be primitives, referring to causation, lawhood, experiences, independently of any structural role they play.  That will be enough to avoid Newman's argument and to pin down the reference of the other terms.  Note that this is broadly in the spirit of Russell's own reply to Newman, where he accepted that certain notions (I recall that he said spatiotemporal copunctuality) would have to be primitive and not structurally defined.

Moises' response:

Returning to the problem, how can one determine which terms will be defined structurally and which ones will be primitive? I suppose mass, charge, etc, must necessarily be defined structurally taking into account the roles they fulfill in physics, but why should lawhood or causation be defined that way? And would defining them that way would require of them something akin to an ontologically platonic status (like taking the objects of set theory, for instance, as fundamental)? And why is such a defense to Newman's argument effective? All of these questions appear to me to be very difficult, and was hoping perhaps you would comment.

I think the general motivations for the Russellian view on the mind-body problem are weaker than the motivations for pure structuralism.  For the Russellian view, you just need the idea that basic physical terms are theoretical terms, characterizing physical properties as the entities that play certain roles without specifying the intrinsic nature of those entities.  That opens the door for type-F monism and the like.  But it certainly doesn't force one to general structuralism.  For that, one would need the much stronger claim that all terms are theoretical terms, or something like that.  I don't see any reason to accept that, and certainly you can't get the stronger claim from the weaker claim.

For example, if one takes a Ramsey-sentence approach to theoretical terms, there will always be O-terms in which the sentences are cast. If one thought one could continue eliminating the O-terms indefinitely with more Ramsey sentences (yielding "global Ramsification"), one would eventually arrive at a version of Newman's problem.   But one might instead take this threat of vacuity as a good reason to deny that all terms are theoretical terms.  And in fact, there isn't any obvious reason to think that all O-terms in typical scientific theories will themselves be theoretical terms and therefore eliminable.  One can't plausibly eliminate logical/mathematical or experiential terms in this way, for example. And presumably there can be other O-terms too, such as "law" and "cause", and so on.

The question of what are the fundamental O-terms is very interesting. On my view it is one of the deepest questions in philosophy.  I have discussed it briefly in various places -- e.g. in Powerpoint presentations on primitive concepts and on terminological disputes, and in much work in progress on the topic.  On my view, the best candidates include phenomenal, logical, mathematical, and nomic concepts.  Some would add spatiotemporal concepts, but not me.

Then there is the question you raised of what is special about the primitive concepts.  I don't see why these must lead to ontological platonism.  But arguably they require that we have some sort of non-derivative grasp of the concept, whereas the other concepts allow a grasp that derives from our grasp of these primitive concepts.  Just what that non-derivative grasp consists in is unclear.  One possibility is the Russellian idea of acquaintance in experience, but maybe there are others.  That depends very much on how one thinks about concepts and intentionality in general.  For my part I think this is another of the deepest questions in philosophy.

I don't think one needs to answer these latter questions to reply to Newman's argument.  The argument is only an argument against pure structuralism.  One can respond by denying pure structuralism. Perhaps there is some further problem for the denial of pure structuralism, but that problem will be independent of Newman's argument.  But of course the question of the status of primitive concepts is of enormous independent interest.

Comments

[Robbie Williams]

I guess the "global descriptivist" (GD) view of how words get their references/extensions that Lewis examines in "Putnam's paradox" (1984 AJP) is one version of "almost pure structuralism". The GD theorist says that facts about what meanings words have is fixed as _whatever assignment of meanings_ would render a set of sentences that constitute the folk's "total theory of the world" true.

A general response to the Newman problem, therefore, is the one that Lewis canvasses in that paper in defense of GD. His basic thought (which needs and gets refinement) is that the Ramseyfication quantifies only over "natural properties". Assigning arbitrary set-theoretic constructs as the extensions to predicates, even if they would make total theory true, just isn't doing what is asked for---we haven't found _natural_ properties that render true total theory.

Suppose that idea can be tidied up and made to work. There still seems a sense in which this view has it that science (and language more generally) is "characterizing the world" in purely structural terms---it still leaves open the intrinsic character of these natural properties, for example.

Maybe the way to describe this response to Newman is as the view that "natural property" is the single non-structural notion required. If so, then the Lewis-style response is still interestingly different from DJC's suggestion, since it requires a "thin" metaphysical notion to be non-structural, rather than the "thick" notions such as "cause" "law" "experience" or whatever.

[Jason Zarri]

I realize this wasn't your main point, but I have a nitpick: Regarding this sentence: " But arguably they require that we have some sort of non-derivative grasp of the concept, whereas the other concepts allow a grasp that derives from our grasp of these primitive concepts".

I think there's an ambiguity there, of a de dicto/de re sort. Are we talking about *a non-derivative grasp* of primitive concepts; or a grasp, *simpliciter*, of non-derivative, primitive concepts? It seems to me that they are logically independant: Depending on one's theory of concepts, it might turn out that the grasping itself is primitive; that we don't grasp the concepts in virtue of doing anything else (which would seem to pose problems for physicalism). And if "non-derivative" modifies the 'grasping' in this way, it seems there is no reason why *complex* concepts couldn't also be grasped non-derivatively. But if "non-derivative" modifies the concepts grasped, the *grasping* need not be basic, and might be reducible to physical processes.

[djc]

Jason: "non-derivative" was meant to modify "grasp". The idea is that for theoretical terms, there's arguably a sense in which our grasp on the theoretical terms derives from our grasp on the O-terms (exactly what sense, I won't try to specify here!). But for the primitive concepts, it's arguable that our grasp on them won't derive from our grasp on any other concepts. Of course that's not to say that it won't derive from anything at all: e.g. it could derive from some underlying physical/phenomenal processes.

Robbie: Yes, global descriptivism seems a sort of pure structuralism. Unlike Russell's and Carnap's pure structuralism, it may allow basic property terms as well as relation terms, but the same issues arise. Lewis's move of invoking natural classes to avoid the problem parallels Carnap's move of invoking "founded" relations to avoid the problem (which he belatedly recognizes in sections 153-155 of the Aufbau, in a secion marked "this can be skipped"!). Something like this is also anticipated by Newman, who discusses the possibility that the structuralist could invoke "important" relations. As Newman, Friedman, and others have noted, this seems to step away from pure structuralism to a view with a single non-structural primitive, in the higher-order property of "foundedness" or "importance". Likewise, Lewis's invocation of natural classes is a step away from pure structuralism. I think he sees himself as providing a second-best alternative to global descriptivism rather than defending it, but that's probably just terminological.

One might see these views as singling out a single primitive concept ("natural" or "founded") that gets its referent nonderivatively, and that can then be used to help derivatively pin down the referents of all the other concepts. That would be a view analogous to mine but with a very thin primitive concept base, as you suggest. Alternatively, one might see these views as giving naturalness a special role in the metasemantics. I.e. rather than assigning a special referent to "natural" and proceeding from there, rather, there is a special constraint on assigning referents to all terms, requiring that these referents be as natural as possible. I suspect that the latter is closer to Lewis's view. Of course one could turn the latter view into the former view by treating the constraint as "just more theory", but Lewis clearly thinks that the just-more-theory move isn't innocent, so he'd probably regard the two views as inequivalent.

I've always been puzzled about why Lewis concedes so much to Putnam, though. "Global descriptivism" seems to me to be a highly implausible view of reference and it's not at all clear why Lewis, or anyone else, should be committed to anything like it, as opposed to local descriptivism for some wide class of terms. (I recall that Wo had a post a while back arguing that Lewis is independently committed to rejecting it!) Putnam's argument seems to require something like global descriptivism to get off the ground. As long as one recognizes that some terms/concepts don't function derivatively, then the argument fails. I suppose Putnam thinks that this will require a "magical theory of reference", but it's not clear why it should be magical: there could be some straightforward metasemantic story to tell about the reference of primitive concepts (though of course one will then have to resist turning all of this into "just more theory"). And as various people have pointed out, Putnam himself seems to privilege the referents of certain terms, such as logical/mathematical terms and experiential terms, so that these have fixed (and apparently nonderivative) referents rather than being subject to arbitrary assignments. So the issue already arises for these terms.

I suppose that for Lewis, a special issue may arise because he can't privilege experiential concepts, as he's committed to a Ramsey-style analysis of them, while he might also want to give a Ramsey-style analysis of all physical concepts, and a reductive analysis of nomic concepts as well! If so, it might be unclear just what concepts are left to be privileged. Perhaps logical/mathematical concepts, but these aren't enough to fix the reference of everything. Perhaps spatiotemporal concepts: I don't recall anywhere whether Lewis is committed to a reductive analysis (e.g. Ramsey-style) of these concepts (?). If not, then these might be good candidates for fundamental O-terms, and could be invoked to avoid the problem.

If Lewis is committed to a reductive analysis of all these concepts, one can see why global descriptivism might threaten, with a potential regress might arise that could only be settled by invoking some further constraint such as naturalness. But one might say instead, so much the worse for giving a reductive analysis of all these concepts. It's not clear what the cost is in taking some concepts in the vicinity as primitive.

[Simon Floth]

DJC: I agree that spatiotemporality isn't "intrinsic", or therefore something we have a "primtive" concept of. But if Newman was right it seems physics would be somewhat vacuous without trading on some primitive concepts.

I think the unholy clue to finding primitive concepts is to think primitively. Q. What goes on spatiotemporally? A. The dynamics of matter. Materiality and dynamism might thus be the "intrinsics" that we have primitive concepts of which are the logical foundation of physics. Granted, mass, velocity and acceleration are quantative, dispositional, and characterised by equations. But are they they core or merely the limbs of the concepts of materialty and dynamism? I suggest the latter; Newton must have some intuitive grip of what he was qauntifying before he got started on the maths or even the measuring, and perhaps even before the conceiving of relevant dispositions.

This is something that Russell, and "type F"s in general, like Gregg Rosenberg (who presents impeccably the case that dispostions are insufficient for consciousness as we know it) seem not to have duly appreciated.

The following remarks are more incidental. 'Intrinsic' is vague. Dispostions to internal motion are intrinsic to a mechanical clock. I don't doubt that we know what we are talking about when distinguishing the "intrinsic" from the dispositional, but we we could do with a nice workable definition of it. (Clarifying primitivity of concept should go hand in hand with this.) I offer one in a just-finished paper which also adds detail to the above arguments and adds a bonus derivative argument against physicalism, which proceeds as follows. Dynamism and materiality, all that physicalists could claim as intrinsic, are insuffcient for certain knowledge we have (and it doesn't concern zombies). The paper is not online yet, but if anyone wants an emailed copy, request it from [email protected].

[paul]

i dont see why Newman's attack is devasting. SR says science describes the structure of reality but not reality itself. or SR states science is a finite set of axioms which describe a domain of reality. this descriptive structure is isomorphic to the informational content of the domain of reality under investigation. however this descriptive structure is not isomorphic to the intrinsic reality being investigated. we tend to think this intrinsic nature to be inaccessible via descriptive methods.

what allows science to describe and predict nature is the potential isomorphism between the informational content of our descriptive theory and the informational content of some domain of reality. in other words, informational content is a result from a structure. to continue further, structure is derived from a metric introduced on some space. the informational space reflected by reality is in sense a hologram. this hologram is not "ultimate" reality but is a homomorphism(automorphism?) between reality and itself. this hologram lends itself to description and is isomorphic to a descriptive theory with the same structure/informational content. (obvious)

the question of course is why is reality itself inaccessible. surely there might exist an isomorphism between reality and our description of reality. yet the view of SR, says this isomorphism is inaccessible to us(the observers) at least.

Newman's argument is simply that pure structuralism leads to there existing an uncountable infinite number of *possible* descriptive structures which are isomorphic to the informational content of some domain of reality. however it does not say there are an uncountable infinite number of isomorphic structures to the informational content of some domain of reality. there exists a countable infinite number of them. which is no surprise, surely there is an uncountable number of falsehoods but a countable number of truths.

[Gregg Rosenberg]

I've thought alot about Russellian ontology, but not so much about which concepts of ours might be primitive.

I do think we have a concept of "matter" that is not structural, and that this concept is roughly the medieval view of matter. It's related to Aristotle's copy theory of perception, in which matter is taken to consist of qualities like those we find in perception.

I would also agree that we have a non-derivative understanding of change and dynamism that we impart to matter.

But I don't believe this view of "matter" is part of our scientific view of the "physical". If anything, at least the qualitative half of our intuitive view of matter has been explicitly rejected and removed from our scientific view of what the "physical" is, thrown out along with the practice and theory of alchemy that had been built upon it.

In the book I did address (albeit cryptically) in a footnote to chapter 13 my nascent view of how structural physical concepts latch onto the world :

After explaining on p. 236 how physical concepts describe patterns, I footnote this: "When making this claim, I do not wish to deny the importance of indexicality (i.e., designation) in fixing reference. Likely, physical concepts contain indexical components, as "electron" may express a rigid designator. As Daniel Stoljar has pointed out, electrons are arguably just the categorical natures that play the electron role in our world. The more important point is that, even if some categorical nature is picked out indexically by these concepts, the indexical place functions much like a variable in the conceptual structure. Even if the value of the index anchors the language system to categorical natures, and even if it does so in a way that depends on the deictic orientation of the concept user within its physical context, it is still functional roles that do the most essential work in fixing the physical category applied to that nature."

Unpacking this a bit: I think at least concepts of causation, intrinsicness, quality, certain indexical-based concepts ("self" and "not self", "here" and "not here"), some topological concepts (e.g., connectness and containment) are grasped non-structurally. I think from these we can bootstrap other composite but still non-structurally understood concepts such as "agency" and "other". This gives us pivot points on which we can begin to hang our structural understandings, so they are anchored non-arbitrarily to the world.

[Simon Floth]

Gregg: I think you are are right about the "scientific worldview", and by implication about the position of physicalism, the name of which is an obvious distancing from the materiality suggested by 'materialism', irrespective of the fact that the two are often used synoymously. For my own part, I think that if physicalism were to reverse its typical evasion of the intrisic, it would look rather simplistically Hobbesian, metaphorically waving a mechanical clock, dubiously regarded by Mary the newly red-seeing neuroscientist. In other words, I think elliptical structural stories are an important strategic diversion from the relevant intrinsics.

My comment about materiality and dynamism not receiving due consideration was more in the sense a broad survey of logical space with a view to what might be intrinsic, not specifically with relevance to debates over physicalism, which, admittedly, is perhaps more your focus.

As to intrinsics and logical space, since the intrinsic is in contradistinction with the structural, I suggest that counting non-identical intrinsics, the different things (if there is more than one) at the base of the great pile of relations, is the only logically non-subjective way to justify the cardinality implied by monism, dualism, triplism, etc.

Such implies that if materiality and dynamism are non-identical and intrinsic, and admitted by a materialist, then the latter is a dualist. It also does a lot to make interactionism appear logically viable, since Newton's laws would then be descriptions of precisely such interaction. The question of what charges and fields are, of course, just gets answered by the fact that these, however causally salient, are only structurally defined.

Finally, if one were to grant those two intrinsics and add just one more, like phenomenal consciousness, they would have triplism. Whoever heard of that? But it seems to me the only viable way to assert that we have a clear notion of what intrinsics known relations involve, which may in the final analysis be necessary for the ultimate grounding of such knowledge of relations.

So there may after all be a neglected zone of logical space, even if Russell and type F monists are not to blame for such neglect, which of course, they aren't, much more than Descartes, Hobbes, Berkerly, Leibniz or Spinoza. It's merely that Russell had his finger on the notion of structure in a metaphysical context, and thus might have been more likely to wander where we just did.

[wo]

I'm not quite sure I understand the relation between the threatening structuralism and type-F monism. Isn't all the 'structuralism' you really need for type-F monism the exceedingly plausible claim that being (actually) spin is a matter of satisfying a certain theorical role, a role that says nothing about the intrinsic character of the property? As far as I can see, the role could still involve all kinds of clauses not only about causation, but also about how spin is distributed in our surroundings, what effects it has on our measuring instruments, etc. I don't see how one could reasonably maintain that our vocabulary for fundamental physical properties is a priori tied to a certain intrinsic character.

A more serious threat to type-F monism seems to be the anti-quidditistic rejection of intrinsic characters, which leads to an interestingly related kind of structuralism: as Robert Black argues in "Against Quidditism", without intrinsic natures, the only differences between possible worlds might be purely structural differences.

Re GD + naturalness, I think naturalness here must be understood as an external, metasemantic constraint, rather than a single primitive term: Holding the meaning of "natural" fixed won't prevent us from construing all kinds of deviant interpretations for the rest of the language. (We could for example render "all predicates denote natural properties" true by re-interpreting "denote" as "denote on Fridays, but not on the other days of the weak".)

I think descriptivism plays a very different role in Lewis's theory of meaning than it does in GD and in your theory, Dave, which is why I think it doesn't matter much for Lewis if there are virtually no primitive concepts. I've tried to explain that, but it got horribly long, and since it isn't really related to the main point here, I've posted it on my own blog. But directly on your questions about terms Lewis took as unanalysable: identity, parthood and instantiation are the only ones he mentions, as far as I know. Spatiotemporal vocabulary might be a good candidate too, yes. And maybe also "perfectly natural". ("Maybe" because with tropes or universals, perfect naturalness will become definable, and also because it seems to me that Lewis actually defines "perfectly natural" in terms of "similarity", "duplication", "intrinsic", "structural", "complex", etc., because these are old concepts we understood before Lewis came along with his new term "perfectly natural" that he explained in terms of the them.)

[djc]

Thanks, Wo. Anyone interested in Lewis's views on these things should looks at Wo's very interesting post on his own blog. I've posted a longish response there.

On the other points here: I agree that one doesn't need anything nearly as strong as structuralism to get type-F monism off the ground. Re naturalness, I was thinking that the theory might build in various first order claims such as "mass is natural" and so on. Presumably that sort of thing will constrain deviant interpretations pretty heavily.

[Simon Floth]

The paper I mentioned above, addressing the intrinsic, primitive concepts and physicalism is now online. I've also just posted one on the potentially surreptitious expression 'what it's like'.

[Damon Woolsey]

Mr. Chalmers,

You stated that a primitive concept "…require[s] that we have some sort of non-derivative grasp of the concept... Just what that non-derivative grasp consists in is unclear."

I wrote a paper for a class last semester that explores one potential explanation for our "grasp" of primitive concepts (which I refer to as 'simple concepts'). The basic assumption is that we are parts of the world, and so our modes of perception and rational thought processes are functions of (some parts of) the world. Furthermore, as parts of the world, we are made of the same "stuff" as everything else, and follow the same natural laws as everything else. Now, I define a simple concept as an atomic concept, not divisible by a rational mind into components or parts. The fact that a rational thought process is a function of the world, and that a simple concept is an atomic part of certain rational thought processes, means that the world must work that way (in the sense of a natural law or principle).

But that isn't quite the whole story, for quantum physics seems to contradict many of our simple concepts. What I call for is a multi-level metaphysical reductionism, where the intrinsic properties of some molecular entity are derived from the inter-relations of its component parts. The intrinsic properties of the component parts needn't be represented explicitly in the intrinsic properties of the whole. Thus, a simple concept as some "way in which the world works" is qualified as being the way the world works at a particular level of unification.

I conclude that there are a great many intrinsic properties of our world that are inconceivable to us (e.g. quantum spin). However, at our level of unification, we do in fact have a non-derivative grasp of simple (primitive) concepts, which are just the way our world works (at our level of unification).

I am aware of the explicit circularity of this view; basically that we know that the world works this way, because our minds, as parts of the world, work that way. However, I think that it is rather satisfying, if we are willing to make the assumption that a) there is a world, and b) that we are part of it.

[Carey Carlson]

Regarding the primitives of Russell's analysis of physical theory, generic events are the individual relata, and these are ordered amongst themselves by the relation "cause-and-effect." From those primitives, all physical structure is to be derived. Let's call this logically concise program "bare-bones eventism." The same bare-bones eventism is conspicuous in Whitehead. Since it hasn't produced fruit in physics up to now, bare-bones eventism lies fallow, considered too weak to bootstrap its way to a full physics, or else forgotten entirely. But I've found a way to construct energy from it, which leads me to expect it will make a comeback soon. In that case, let us ask about the "grasping" of the two physical primtives, "events" and the "causal relation." What "grasp" did Russell or Whitehead exhibit of their own two primitives? The "least hazy" partial answer involves their agreement on a substantial point: they both said that the relation of cause-and-effect is the selfsame relation as before-and-after. There is only the one physical ordering relation, and we ought to revise our split intuitions of "time" and "causality" to the point at which the divergent notions are conflated into one. They held to that opinion because of their mutual footing in bare-bones eventism, and their respect for Occam's Razor. Their formal intuitions about relations and structure, and their realism in regard to eventism-as-physics, meant that a single ordering relation was responsible for the constitution of the physical world. The logical character of their eventism thus "recoils" upon their grasp of the physical primitive. The grasp of the "event," the relatum, varies with the grasp of the relation. And from here, Russell and Whitehead trail off in various directions, with mixed results. In any case, bare-bones eventism, as a theory with only two primitives, is a good object lesson in which to discuss the grasping of primitive concepts, because of its minimalism. Also, it is alive today in the work of Rafael Sorkin, under the name "causal sets." He has formulated cosmology in its terms, contributing to the theory of Hawking radiation and precise values of the Gravitational Constant. Eventism is back!