On a side note, I just found out that a recent paper of mine partially reduplicates an argument made by Gualtiero in one of his papers (with some differences in emphasis). Nice.
At anyrate, I have been thinking about multiple realization. Now, whether these kinds of arguments refute any of the families of Identity theory is one question. This, I suppose, was the work it was originally intended to perform. I think that it does not even do that since we can formulate a type/sub-type distinction such that distinct realizations of, say pain, P1, P2, and P3 can be seen as pain subtypes. The type/subtype relationship is already robustly applied in biology so applying here would not be a stretch. But either way this is besides the point I want to make now.
Intuitions about multiple realization have morphed from intutions about a particular type of physicalism being unintuitive to physicalism itself being unintuitive. But this move seems problematic to me. Suppose that the physicalist did admit that mental types were multiply realized. What this will mean is that Pains, say, could be instantiated by many different physical states. The idea, then, is that it is only functional orginization, or causal connections, or some such thing, that matters and so even non-physical properties/states could realize mental properties. But why think this? What reason do we have to think that non-physical properties or states can do anything like what is required on either front?
Why Multiple Realization Arguments don’t Refute Physicalism
76 Comments
Comments are closed.
Richard,
I posted on something like this idea of going type-subtype a few weeks back under “Specificity and Multiple Realization.” In fact, I’m hoping to finish a draft of a paper on this topic around Oct. 1.
So, are you saying that Multiple Realization arguments against physicalism don’t work because they have some counter-intuitive consequences regarding what sorts of entities can instantiate mental states?
Hi Ken, thanks for the pointer. I read the original post but did not see the subsequent discussion that ensued.
So, then do yo take yourself to be offering an argument against this move in that post? I mean, do you see the appeal to grain and specificity as the same kind of move that I suggested by type/subtype? It is not obvious to me that it is the same. I was pretty much convinced by our last discussion that the leveled view you and Gillett have worked out is plausible and so I think psychologcal may be multiply realized in the way you suggest, but my suggestion was this being true is not problematic for physicalism since the various realizations can all be seen as subtypes.
Hi Charlie, thanks for the question.
Yes that is exactly what I am saying. It is completely mysterious how non-physical properties could have any causal powers or how non-physical entities could instantiate functional orginazation of any kind whatsoever. People usually just assert that non-physical minds could instatiate functional orginazation, but this is absurd! Or at least, to be fair, it stands in need of an argument…
Richard,
I seem them as related, but exactly how is not clear from a brief post.
So, it looks as though you might have in mind more than what I put in my original post. The point there was that introducing subtypes might better your chances of unique realization, but it does not establish UR. Here, it seems you might be saying that any time there is an apparent difference in realization of a type T, then you can create new subtypes ST1 and ST2 which will be uniquely realized. That would be different than what I posted, but I think it won’t work. The basic problem is that you’ll have to turn to kinds of subtypes that you don’t typically find in science. How that works in greater detail is part of what my paper is about.
Incidentally, I think Oron Shagrir makes something like this move, or alludes to something like this move (its only like one sentence), in his “Multiple realization, computation and the taxonomy of psychological states”. And I think Bechtel and Mundale has about one sentence hinting at this as well, only I can’t find it.
But, I’d be interested in seeing your paper, if you are circulating it.
In “Neurophilosophy at Work” 2007, second essay, page 26, Paul Churchland says:
“Once again a diversity of material substrates does not entail diversity in the underlying laws that govern those diverse substances”
On the other hand, Churchland move is an astute move to guard the truly correct intertheoretic reduction (which seeks to find a unified theory for all science) despite the fact that it is possible that mental capacities can be realized in dinstinct substrates, only that independently of the substrate in which a mental capacity is instantiated, it have to be govern by physical laws (physicalism)
But if we take into account our current state of knowledge of neurophysiology and anatomy, not only functionalism (and its jewel of the crown arguments: the multiple realizability arguments) seems to be false, the idea of multiple realizability of mental capacties is false true.
In our biological brains, structure and function are metaphysically inseparable, and if we want to simulate, reproduce our biological intelligence, we have to take care of the exact matter arrangement.
This is not to say that the initial propects of AI, or machine consicousness are vains, just that they could create better intelligence (with greater computational capacity, memory storage…) but not noisy biological intelligence.
Thanks for the response Richard. I have a bit of a concern about your analysis of multiple realization (MR). It seems like there are two ways of going about understanding MR arguments against type physicalism.
The first is to accept that we’re working inside a physicalist (or “naturalistic” if you prefer that buzzword) ontology and to run MR concerns against type physicalism. I’ll call this “restricted ontology MR” (ROMR).
The second is to accept that we need an ontology of some sort (a no-holds barred sort of approach to ontology where anything goes!) and we run MR-arguments therein. I’ll call this “unrestricted ontology MR” (or: UOMR).
In order for your concern to get off the ground, we have to endorse UOMR (ROMR outlaws non-physical anythings by hypothesis). So, why would anyone accept UOMR? Why can’t I defuse your reductio against MR by endorsing ROMR? That would let me say that there’s no such thing as a non-physical mind and avoid what you take to be absurd.
As for an argument in favor of ROMR over UOMR, that just comes down to an argument in favor of physicalism. After all, the only different between ROMR and UOMR is that the former endorses a physicalist ontology while the latter doesn’t.
I’m a jerk because I submitted the comment prior to finishing what I wanted to say. So, sorry for any confusion!
So, I was saying that running multiple realization arguments inside a restricted ontology (ROMR) avoids your concern about the absurdity of MR arguments against physicalism. Of course, that simply proves your point.
What I wanted to get at was that I understood MR arguments to be an argument against type physicalism, not physicalism simpliciter. So, running MR arguments inside a physicalist ontology isn’t just about making room for those with dualist predilections. (Question #1) Do you suggest that MR arguments have absurd consequences, need to be junked, and type physicalism should rule the philosophy of mind landscape?
Now (to get to a second point), as for how non-physical entities could have instantiate a functional organization, could someone argue that a non-physical entity N has a definite functional organization F. For N to exemplify F, N simply has to have inputs and outputs correlated in such-and-such a way. These inputs and outputs can be non-physical; so, I can understand what it means for N to instantiate F–N instantiates F just like a physical system P instantiates F. It’s just that what gets correlated in N are non-physical inputs and outputs while what gets correlated in P are physical inputs and outputs.
Now, don’t ask me what a non-physical input or output is because I don’t know. But, that concern is subsidiary to the fact that N correlates them; and, it’s the correlating of the non-physical inputs and outputs that lets us say N instantiates F.
(Question #2) Could someone reply to your reductio as follows? (summarizing what I said above): The MR-supporter could respond to your concern by admitting that she doesn’t know what a non-physical input or output is; but, that doesn’t matter. What does matter is that these non-physical inputs and outputs are correlated in such a way that the non-physical system is best characterized as instantiating a functional organization. Hence, non-physical systems can instantiate functional organization; it’s just not quite clear how we understand the natures of the inputs and outputs.
Hi Richard, I guess I am not tracking the concern, so let me throw out some stuff and see where we land. (I hope this doesn’t repeat earlier points, but gets to some similar places from different directions).
First, the idea that MR undercuts ALL versions of physicalism is prima facie bizarre when we recall that one version of physicalism, so-called ‘non-reductive physicalism’, is founded upon realization and multiple realization.
Second, why is anyone working with intuitions? Who cares about people’s inutitions about what the natural world may or may not be like? The sensible defenders of non-reductive physicalism have looked at scientific findings and argued that, given our mechanistic explanations from an array of sciences we are now justified in accepting that all entities loacted in space-time are composed by the microphysical — in the case of proeprties this would be the claim that all properties instantiated in individuals loacted in space-time are either realized by, or identical to, properties of microphysical individuals. (See Part 3 of melnyk’s PHYSICALIST MANIFESTO for the best presentation of such a defense).
So, where are we? MR underlies one theoretical framework for physcialism; it is claimed empirical evidence supports all properties in space-time being realized by the microphysical properties. Now someone has an intuition about (i) what a physically unrealized individual(s) might have with regard to properties; and (ii) whether such an intuited non-physical individual(s) could realize mental properties.
Well, I guess I am a little dubious whether we should trust such intuitions at all — people have not done well with intuiting the nature of the kinds of individual and property found in space-time let alone something this unfamiliar. And, in addition, even if such intution were reliable why would it not be simply trumped by the putative empirical evidence? How does the intuition here help at all?
I am going to post in a different message about the sub-types appraoch, best, Carl
Hi Richard, I am not completely sure about what you mean about the sub-type strategy. Is that the idea of focussing not on hardness say, but the narrower property (the sub-type?) hardness-in-human-teeth? If it is the strategy, I guess I think it faces the following two worries. Curious to see what you think.
Start at the beginning to see we are on the same page: Writers like Putnam and Fodor lacked any precise framework for the ‘realization’ of properties or scientific composition more generally. But Putnam and Fodor could none the less use evidence for such multiple composition to undermine the Nagelian model without having any account of scientific composition. For “Being A is identical to being B’” is not true in the many scientific cases where we have multiple composition due to the failures of coextensivity that result between the higher and lower level entities.
Narrow identity response (sub-type response?) are arguably the most common response to the Putnam-Fodor critique, found in early work of Lewis and Kim and still defended by a number of writers. The suggestion is that one can get so-called ‘restricted’ or ‘narrow’ identities, such as so-called “species specific” identities in the case of psychological entities, focused upon a narrower kind of higher level property – what you appear to call a ‘sub-type’. Consider the property of having a certain Knoop hardness index – it is multiply realized in both teeth and metal fillings, being realized by the different relations of the atoms/molecules found in these two substances. In response, the restricted identity (aka sub-type) response seeks to identify the narrower sub-type (?) hardness-in-human-teeth to such a lower level property. For the narrower higher level property may have a unique realization base and hence is always found with the same lower level properties. Thus the Putnam-Fodor obstacle to identities is avoided and one can hope to get identity claims (even Nagelian reduction!) back on track.
OK, is that the basic approach? If it is, here are two worries. I get a little agitated pushing the first one, so I keep it short…
First worry: Do we really want to say that as well as the property of having a certain Knoop hardness index, as materials science suggests, we should also accept that a tooth also has the property of hardness-in-teeth or whatever analog is proposed? That looks perilously close to being ad hoc, maybe it looks less so in psychological cases. There is no guarantee our empirical evidence makes the existence of such sub-types
But there is no guarantee our empirical evidence makes the existence of such sub-types plausible – you need to make such a case example-by-example. Otherwise it looks like the usual analytic theft rather than honest naturalistic toil…
The second worry follows in another post (busted the character limit)…
The second worry (cont)
Second worry. I will give you the sub-type property, for argument’s sake (and despite its effect on our blood pressure), and further assume that we do have a unique, rather than multiple, realization base for the relevant property, in this case the property of hardness-in-human-teeth. None the less, one still does not get identities. Here is why, breifly put. Scientific realization relations are asymmetric, many-one and relate entities of qualitatively different kinds, among other features. In contrast, identity is always symmetric, one-one and relates entities of exactly the same kind. As a result, given the differing natures of compositional relations, and the relation of identity, it is plausible that one will not get restricted identities even when one does have coextensivity through a unique realization base. To note just some of the obvious problems: one instance property of hardness-in-human-teeth cannot be identical to the many lower level atomic/molecular properties and relations that materials science tells us would be its unique and coextensive realizers. (And of course the differing nature of the realizer properties/relations (in certain bodning) and the character of the uniquely realzied proeprty, in hardness-in-teeth, preclude identity — qualitatively different entities cannot be identical…).
These are completely general features of scientific realization which will apparently apply in the psychological cases. Thus multiple realization is plausibly not necessary to block identities between higher and lower level properties in the sciences, since even unique realization suffices to block such identities given the nature of scientific realization. The sub-type strategy thus fails anway — so why bother with the abominations of th sub-types? (I make this kind of point briefly in my 2007 J. Phil paper and at greater length in a book manuscript I am finishing-up).
So, if I got what you mean by the sub-type approach right, any ideas about responding the these two worries? Best, Carl
Hi, Richard,
It’s not obvious why the concepts of physical and causation have to go together. Why couldn’t there be a world at which nothing is physical (that is, none of our physical properties are instanatiated) but things cause one another?
Tony
It doesn’t seem ad hoc to me. It is fairly standard neuroscience to restrict oneself to a particular species. Papers are usually of the form ‘Escape response in the yellow-bellied jackal’ or ‘Visual representations in macaque V1.’ People certainly hope that the results will extend to other species, or subsystems, and this is something that is determined empirically.
Is anyone an identity theorists with respect to digestion? This isn’t a rhetorical question. Why not study something that is well developed rather than something in which the science isn’t yet settled?
You should probably use the first person singular when talking about the blood pressure with explanatory restrictions: they are fairly ubiquitous.
Your second point though is interesting. It helps to point out just how strong an identity claim is, and to provide an alternative hypothesis about explanations in biology, about how they don’t satisfy the criteria for identity. However, to just say that scientific
realization relations are asymmetric, many-one etc would be to beg the question without a specific example (and that would just show you are right with that example–in biophysics I bet there are lots of identity claims, e.g., ‘The gna variable in the HH equations is identical to the conductance in sodium channels.’).
Nice to see some discussion developing here. Sorry I haven’t been a part of it. My internet connection mysteriously shiyt down two days ago and only cam back up today (luckily since the technician wasn’t comming until two weeks from tomorrow!!)
Anyway, I am going to start reading and thinking and hope to reply in a bit
Expresssions like this seem to me to be ambigious, at best. One could read them as offering “property P as found in species X” or as offering “property P-in-species-X”. But, in this post, I think Carl is understating just how ad hoc the categories will have to get. For one thing, if you look to how traits vary in a population, you’ll have to cut things up even more finely.
So, take, say, human color vision. Things are already split into various forms of anomalous trichromacy and dichromacy, but don’t you think it ad hoc to come up with a different psychological property for human color vision for each of the possible combinations of photopigments in the cones?
So, for example, it now appears that different people have different numbers of red genes (conservatively four) and different numbers of green genes (conservatively seven) and that these copies can differ among themselves. Should we make dozens of different human color vision categories to tract all these combinations? That sounds like a philosopher’s trick to be sure.
We should study each category empirically to determine whether there are indeed different types at the psychological level.
However, I understand your point. Not all genetic differences will yield psychological differences. Whether this means that there is multiple realizability, I’m not sure. If the genetic differences yield no differences in the electrical properties of the cells (and if they did there would probably be psychological differences), does that mean we have multiple realizability? On the surface, yes. But let’s press this. Perhaps the reducing claim could be that pscyhological property X reduces to (or is identical with) electrical property Z. So even if Z is multiply realized, that wouldn’t mean that X is.
It’s like temp reducing to mean kinetic energy of molecules. Because this doesn’t specify which type of molecules, that might mean that temp is multiply realizable (e.g. the temp 4 degrees kelvin can be realized in Helium, Argon, etc). Or it just means it is reducible to mean kinetic energy, which is not reducible, but that doesn’t matter.
I’m not sure how I feel about this, honestly. Just playing in possibility space.
.
Absolutely. Carl and I are 100% behind that. But, we also think that much contemporary science looks pretty promising for MR.
Ah, but this is what’s great about the cone opsins. The different genes correspond to different proteins, which typically, but not always, have different light absorption spectra. The difference spectra give rise to distinct color perceptions. So, they do make for psychological differences. I left that part out.
And the variability in cone opsins is just the smallest tip of the iceberg. There are many, many other components in the human color processing system. They too can vary. So, you can’t, barring philosopher’s tricks, track them at the psychological level by hypothesizing more subtypes.
Is it a philosopher’s trick to distinguish dichromats from trichromats? That, after all, is a simple genetic difference.
I would think not, since it’s a pretty standard distinction in vision science. By contrast, there is, at least to my knowledge, no scientist who wishes to postulate dozens of psychological types for the dozens of different combinations of cone opsins.
If each combination of opsins produces people with different perceptual abilities, I don’t see why anyone wouldn’t want to categorize each one. Indeed, that would be a guaranteed paper, probably in a halfway decent neuroscience journal. At any rate, I don’t see any principled reason to treat this case as different from the tri- vs dichromats. Just that the latter is already used isn’t a compelling argument.
We certainly like to split mouse strains depending on very simple genetic differences (often a single gene is different). You’ll read about phenotype X in mouse strain Y (where Y is different from the wild type by tweaks at one allele). No way you would publish without specifying what mouse strain you are in, it’s sloppy science.
I’m not sure what rides on this, but I think this line of attack against subtyping seems weak, while Carl’s second argument seems more promising.
Apropos of I should ask again: is anyone an identity theorist about digestion? Is digestion identical to some molecular/biochemical process in the digestive tract? What restrictions/qualifications would be needed to make it acceptable (if any)?
Well, if one is giving a descriptive account of the concepts found in science, it does matter whether some concepts are actually used or not. And what I’m saying is that when one consults the science, one does not find both a distinction among all the different cone opsin combinations at the neuroscientific level and a corresponding division at the psychological level. What one finds is that many of the combinations lead to folks being grouped together as having normal color vision.
Neuroscientists have been interested in documenting the variability in the cone opsins, but I’ve seen no attempt to provide psychology properties that correlates to match that cone opsin level variability. Psychologists will note the ways in which cone opsin variability lead to perceptual differences, but they do not hypothesize new properties.
But, as I suggested, the cone opsin variability is just the tip of the iceberg. For one thing, recall that phototransduction involves a biochemical pathway with lots of components, e.g. G proteins, phosphodiesterases, voltage gated ion channels, etc. Each of those too realize color vision and each of those to will likely vary, so there are many more components that are relevant to color vision that vary. Maybe you can swallow the idea that psychology tracks variation in the cone opsins by postulating a new property for each combination, but it’s much harder to suppose that psychology is going to track all those changes. It is one of the virtues of the Dimensioned view of realization that one notes that there are typically many lower level properties that realize a given higher level property.
Now, you could want to urge that psychologists change their ways. But, one still has to address the problem of he move being ad hoc. Recall that being ad hoc is making some theoretical posit just to save one’s view. So, if the only reason that one introduces subtypes is to block multiple realization, then that just is what it is to be ad hoc. So, one needs some independent motivation to split a type into subtypes. This also goes back to the idea of whether or not something is ad hoc depends on what is currently in play in the science.
At this point I have a half written paper on this and I don’t think its ready to be seen…but after working through all the commments here I may be able to spruce it up…if you’re still interested I would be happy to let you look at it!
Richard,
That would be fine. I have about a half-written paper on this too. So, maybe we can exchange in a month or two.
Hi Charlie, maybe I misunderstood the point of your question before. I am not trying to offer a reductio againt MR. I was merely pointing out the people seem to assum that functionalism is compatible with dualism, which seems bizzarre to me. The reason for this is UOMR, as you note, and I agree that what MR shows, if anything, is only that we should adopt ROMR…the point I want to ciritisize is the claim that MR intuitions support the claim that mental states could be realized by non-physical minds.
Hi again Charlie.
Yeah, I think that what you say is basically what people have in mind when they think that non-physical things could instantiate functional orginazation, but the picture you sketched is incoherant. What is a non-physical entity like? What are its identity conditions? I mean, even in broad terms, what are they? No one has an answer. How could such an entity have a definite functional orginazation? What would that even mean? I can tell what it means for a physical system to have a certain functional orginzation (well, there is a debate about what exactly we have to appeal to here, but all of the cadidates are physical entitties that are fairly well understood). So, the problem is much deeper than that of trying to understand what a non-physical input or out put is (though that is certainly tough enough!). The problem is trying to give a coherent accouont of what such a non-physical entity is like or what laws it obeys in anything like the level of detail that we can for physical systems. I claim that we have no idea, none, what such an account would even look like in broad outline, let anlone in the level of detail that physical theory is worked out to (which isn’t very detailed).
Hi Carl, thanks for your comments, they are very helpful.
Re 1: I agree that there is some weirdness here. Part of my puzzlement comes from trying to understand non-reductive physicalism. What kind of non-reductive physicalism is actually supported by multiple realization intuitions (just granting for now that there really is multiple realization of mental states, even along the lines that you and Ken have argued for)? Well, not the kind that endorse a brand of functionalism which holds that non-physical entities could instantiate mental states, or even that non-physical properties could play functional roles.
Re 2: I very much agree with what you say here. All I mean to be doing was castiung doubt on one set of intuitions: intuitions about non-physical entities instantiating functional properties.
Hi again Carl, sorry I got distracted by McCain announcing his VP pick…a shocker…
Anyway, I will leave joining the discussion on the first objection for another time and focus on this one. I am not sure I see what this worry is supposed to be. I am sure it is just me…but let’s say that seeing red turns out to be a certain pattern of synchronous neural firing (as I think). In order to have that kind of neural firing you need neurons arranged in various ways. The various ways may vary and yet the same pattern of synchronous firing may be exhibited. So I then say ‘seeing red is such and such neurons firing in such and such a way. This is supposed to express an identity that holds true for all creatures with neurons in worlds with laws like these and it is symetric and on-one. Now suppose we find that we find some kind of creature that does not exist primarily in the solid state, as we presumably do, but spends its time in a plasma state. In such a state heat is not mean molecular kenetic energy, since the particles cannot move quite as freely as in a gas (the same is true for solids and liquids). What we use instead is some analog or derivable property. So too in such a creature perhaps the pattern of electrical activity is instantiated in plasma equivelents of neurons. IF such a thing happened then we would have to retreat to the red-in-Earthlings/ red-in-plasma-aliens (which is like your hardness-in-teeth/-in-metal) example, and so far we have no evidence that this is true) line that you worry about. But
I don’t care if we call them types or not, but if there are psychological differences here that haven’t been studied, they will be. It’s an obvious set of experiments that someone either has done or will soon do. I guess I trust my judgment here more than the contingency of what has been experimentally established thus far. I know as a reviewer of neuro manuscripts, that would be into J Neurosci, perhaps even Nature Neurosci if they found clear differences. Note I know little to nothing about this field, so for all I know the experiments have been done and there are no psychological differences.
If there are differences, then in careful psychophysics experiments, they will have to start giving the genotype of the subjects, just like now they always say whether they are using di or trichromats. Again, this is just standard in the mouse, where the genetics is better developed.
Maybe you can swallow the idea that psychology tracks variation in the
cone opsins by postulating a new property for each combination, but
it’s much harder to suppose that psychology is going to track all those changes.
This is exactly what psychology will do if there are psychological differences. Why would a psychologist not want to keep track of psychologically relevant biological differences? Or at least a psychologist interested in biological explanation.
I understand you can argue that what we typically (used to) describe as ‘normal color vision’ can be broken up into a bunch of different subtypes with lots of variability b/c of opsin differences. Does this mean ‘normal color vision’ is multiply realized? Or does it mean ‘normal color vision’ has fractionated? Is ‘memory’ multiply realized because there is episodic and procedural memory, or are there two different types of memory? Is there a fact of the matter here, or perhaps just feelings about the appropriate level of fine-grainedness?
Rather than argue about whether there is a fact of the matter here, which would be sort of boring, Carl’s second argument is better and gets at the crux of the matter.
Incoherence is way too strong a word. It’s hard to imagine, but that doesn’t mean it’s incoherent.
I could imagine a bunch of gods interacting with each other. Nonphysical mental interactions. I can imagine a God, outside of space and time, supporting the existence of the universe at every moment. I can’t give a lot of details because I tend to describe things using spatiotemporal metaphors. And I’m not sympathetic to such ideas, but I assume a seasoned Christian apologist philosopher could do some good word ninja here.
Well, the discussion did start out talking about type/subtypes.
But, you are right that neuroscientists and psychologists study how the neuroscience variation leads to psychological variation. And the simple version is that sometimes it does sometimes it doesn’t. But, even when it does, the response has not been to postulate new subtypes, contra Richard’s suggestion.
But, I still like this line, because
1) The view Richard is pushing seems to enjoy a lot of prima facie plausibility. Many people seem to find it “intuitive”. So, it would be good to undermine a view that folks find plausible.
2) The line illuminates what is going on in an actual case, and
3) It helps illuminate or illustrate some features of the Dimensioned view of realization.
So, while I think Carl’s second objection is fine, there seems to me to be some room for some careful work on this other line.
Only one thing to say here about the relevant “type of non-reductive physcialism”.
I actually DO think it allows for the possibility of a non-physical substance with non-physical properties that realize mental states.
The position in some broad sense does allow that. Its holders usually we have any capacity to know whether such possibilities are real.
More importantly, they make claims about the ACTUAL world: That all mental and other properties in the actual world are realized by, or identical to, the properties of physics. But that leaves open what could have been if different substances existed.
So I differ with you on this first point… Though I do see what you are after, some argument that the non-physical cannot have the right nature to causally act in a way that allows it to do the realizing. But I either am dubious we can decide about that, or speculate that very different substances than those we know, if they existed, could have properties that could realize many properties in our space-time. But I tend to the agnostic position… Best, Carl
Hi Eric, well I gave you one example of a multiple realization, but I did rely on my deeper views about scientific realization and I do have a theory. I develop that account in a bunch of papers — I say a few relevant things in my JPhil 2007 paper and more in some forthcoming peices. (Plus the papers I cowrote with Ken work through a bunch of examples that support the asymmetry claim).
The basic point, I *think*, is obvious from mechanstic explanations. I claim the following feature is general with such exaplanations and the realization relations between properties they deploy to drive their explanations.
So, consider our mechanistic explanaiton of why a diamond scratches glass in virtue of its hardness. The explanation is based around carbon atoms and their covalent bonds and the molecular processes of breaking bonds between glass molecules. The lower level processes together result in the higher level process of scratching, but not vice versa. Similarly, the lower level properties whose manifestatons ground those lower level processes together result in the higher level property of hardness but not vice versa.
That is far too quick and sloppy — but do you agree that the relations between properties broached in this and similar scientific explanaiton are asymmetric ontological relations? If you do then that is the kind of feature we find in all such explanations and hence underlies my claim about the asymmetry of realization relations.
If that didn’t help I refer you to the papers mentioned above 🙂 best Carl
But, even when it does, the response has not been to postulate new subtypes,
How could you tell if a scientist is talking about new subtypes? It can’t be the fact that a group of scientists talk that way, as few people talk this way outside of philosophy. What facts of the matter decide the ontology of subtyping?
In neuroanatomy there are lumpers and there are splitters. They agree on the facts of the matter, but disagree on the number of types of neurons (even though they don’t typically use that language). At conferences, they typically laugh about how silly it would be to argue about who is right. That is how I feel about this.
<i>1) The view Richard is pushing seems to enjoy a lot of prima facie
plausibility. Many people seem to find it “intuitive”. So, it would be
good to undermine a view that folks find plausible. </i>
Scientists typically restrict their analysis to species, or to genus, or even to strains of species (as I do with mice in my work). That is why what Richard is saying has some initial plausibility: it’s what scientists do, and for good reason (I don’t take “what scientists do” as authoritative, as they often do stupid things, but in this case it is smart). Again, whether this has any implications for identity theory I frankly don’t know.
Obviously Richard’s view would undermine something you and Carl believe in. I haven’t seen a good reason to think anybody is right so far (including Richard). Perhaps a weakness of the blogs. What paper is it that defends this ‘Dimensioned’ view of realization?
I agree it is too quick and sloppy. Perhaps you or Ken could post a link to a paper that makes the case about asymmetry with a well worked out example (from biology).
Just to be clear: I saw your example with hardness, but argued that in biology we restrict our explanations to particular species (or even strains of species) all the time. So why should my intuitions be violated here? This is just responsible science. Extending explanations, even beyond individuals, is not done without replications and statistical considerations. That’s why we use an N greater than one.
Hi Eric, a couple of different issues are mixed-up here.
First, I think that Ken and I both agree, and have each consistently contended, that it is a case-by-case issue what kinds of entity we should say exist — where that is largely fixed by the empirical evidence. (Although what background ontology is not fixed in that way: one should not just take every predciate to pick out an entity, but that is another story). So we agree that one has to LOOK AT THE CASE and its details. We have also stressed, as you are, that working out what kinds we find in a scientific case, even one with mature, detailed accounts, takes a lot of work.
What we are both pushing on is the too cheap responses that do not bother to do the empirical work. Just because there MIGHT be sub-types shows nothing at all — one needs to work for that claim.
The second, background, issue concerns the general nature of what I term ‘compositional’ relations in the sciences. There is debate slowly starting over that issue. Unfortunately, until recently the procedure has been to take the competing ‘functionalist’ frameworks developed in the philosophy of mind for the mind-body relation and apply them to the scientfic cases. One debate is over whether that works, I contend it does not. However, a growing body of work seeks to offer frameworks directly focussed on the scientific cases and their relations, usually ‘realization’ between properties. Thus Shapiro, Wilson, Polger and Shapiro, myself, Ken, myself and Ken, all have work pressing positions in that debate.
How one answers this general issue can impact the issue over sub-types — especially whether one gets identities. Why? Cos if compositional relation are usually many-one, then even using sub-types will not secure identities between higher and lower level entities. Or so it appears.
My dimensioned framework is partially outlined in my 2007 J. Phil paper, in some forthcoming papers/book, and in my co-authered papers with Ken (where the latter are on Ken’s website right now). Best, Carl
Thanks. I’ll take a look.
I’m taking my cue from the examples given, e.g. dump talk of pain in favor of human pain and octopus pain, dump talk of memory in favor of short term memory and long term memory. (Thanks for bringing this latter example to mind.) I’m assuming, perhaps erroneously, that what Richard and others are proposing is that scientists will adopt this language of genus and species. Then, my first move is to note that this is not actual practice. Given the chance to split, say, anomalous trichromat into further subcategories on the genus-species model in the face of lots of variability in the stock of cone opsins in different humans, they have not done this.
I think that what Richard is proposing is not what scientists do. It only looks that way superficially. Scientists, of course, often do recognize subtypes plenty often, but what Richard is proposing is, at best, carrying this to extremes. He apparently proposing that anytime one finds an apparently difference in realization, postulate new higher subtypes. That’s a lot stronger than what scientists seem to me to do.
Here is a biological example that is, I think, just like Carl’s hardness example. Take the property of maximally absorbing light of 532 nm. This is the putative type. There is experimental work that gives this value for two distinct photopigments both of which are found in the human population. So, what are we going to get for our two subtypes? Maximally absorbing light of 532 nm in photopigment A and maximally absorbing light of 532 nm in photopigment B?
There appears to be nothing in the science of photometry to warrant this creation of subtypes. It appears to be an ad hoc move to block multiple realization.
Ya’ shoulda mentioned the potassium channel example: it is a very good example because it is relevant and the science is relatively well worked-out. I just looked over ‘Multiple Realization and Methodology in the Neurological and Psychological Science.’
I like that you disentangle multiple realizability from constraint
issues (the big killer for early functionalists was their view that we
can get the correct functional decomposition without looking at
mechanisms–we see how well this worked for digestion).
The stuff on psychology is not all that compelling, partly because the
science isn’t close to being well worked out yet, decades if not longer
away from the level of understanding of K-channels, so I won’t even
touch dendritic spines and its implications (or not) for HPMR (human
psychological multiple realization).
Here are some confusions I have with it.
1. MR is too promiscuous using your “dimensioned” view
Before starting with meat, in some sense, multiple realizability is trivially true–vision is multiply realized (there is photopic and scotopic vision, for instance), memory is multiply realized (we have declarative and procedural systems). Potassium channel is also trivially multiply realized (you can have voltage-gated K channels, Ca-gated K channels, K-channels that are always on, etc, obviously different species that fall under the K-channel genus). I think you pointed this out, or someone did here recently.
In your ‘Dimensioned’ view (why did you pick that name?), I’m not sure the above ‘trivial cases’ would count as MR (Multiple realization).
Your definition is as follows (I simplified the original some so it wouldn’t give normal people seizures):
(Dimensioned
Realization) Properties F1-Fn realize
property G in an individual s if and
only if s has property G in virtue
of the powers together contributed by F1-Fn to s but not vice versa.
Using this definition, the property of being a K-channel is realized by the property of being a voltage-gated K channel. So the definition implies lots of trivial cases of realization when we have an obvious genus-species relation.
It would probably be nice to add something to exclude such cases. Using this definition, the property of being a human is realized by the property of being a man. It seems some additional constraints are required so we don’t end up with triviality, where superordinate categories are (by logical necessity) realized by instances of subordinate categories.
For a more relevant example, the NMDA receptor is composed of A and B subunits, which are expressed in different proportions during development and this has functional consequences for plasticity. So, is the NMDA receptor multiply realized because there are these different subunits? This is an interesting case, as NMDA channels are subtyped
explicitly by researchers according to the different compositions of subunits they have.
Using your definition NMDA receptors are realized by different subunit types. It seems your criterion will end up lumping together things that should not be lumped together, things with very different functional profiles (e.g., ion channels are realized by sodium channels) and which are not lumped together in science, except in the trivial sense of being lumped together as subordinate members of a category.
2. Consistent with identity theory?
Let’s pick a case even more sympathetic to your view. In K-channels, does the ability to swap in functionally identical amino acid residues into a K-channel imply multiple realizability is true? I think it does, in your definition. But someone could accept this and still believe in the identity theory of K-channels, no? Could someone say that the K-channel is identical not to amino acid sequence, but the tertiary structure of the sequence and the location and magnitude of certain functionally important charges? Identity theory and multiple realizability are both true, in this argument.
So, while I like your example from the paper with Kenneth, it deserves a full paper-length treatment to address obvious questions like those I asked here.
Or at least that’s my first impression.
I admitted above that this would be a case of multiple realization. If there is nothing there that will create different responses at the higher levels, then we have a case of MR.
We go in circles again with the opsins. I’m asking you what criteria you would use, personally. What reason is there to take di versus trichromat seriously as a split, but not those with two different opsin profiles that display different psychophysical curves? What is the principled reason here? Just saying “Well, that’s what the scientists have done so far” is not an answer to that question. If you insist on that, then I ask what reasons you think they have?
My argument is that there is no principled way to distinguish one fineness of grain from another in this case, that is, there is no fact of the matter here (and I used neuroanatomy as another well know example, and in that case it is well-known to practitioners).
I see two reasons to argue for a distinction. One, there is variability within experimental procedures. That would be principled, but we aren’t talking about that. Second, the interests of the experimenter. This isn’t principled as much as subjective.
Frankly, I think little to nothing rides on this, and was simply balking originally at the original claim that restricting claims to subtypes is philosphical jive. That was an overstatement, at best. It is exactly what we do in science. All the time. That is why science kicks philosophy’s ass: we deal in concrete particulars, and study the shit out of them and don’t assume things will generalize.
Of course, this could be taken to an extreme, only studying one animal your whole life, or something kooky like that (or, not so kooky, as Pepperberg with her parrot Alex showed). But the original claim by Kenneth or Carl or whomever was that you wouldn’t want to restrict explanations to species. Yes, you would. So we could argue about opsins, but ultimately I think my main point is sound, and fairly obvious.
Hi Eric, thanks for reading the paper and for two interesting comments. I think I agree with both at root, but I do need to qualify my agreement.
OK, the first worry was that the definition of Dimensioned realization allows in too many case of MR. Larry Shapiro has pressed a different version of this worry that, overall, we find too many case of MR under our view — so many that this is unintuitive. Ken and I reply in a couple of papers that we do not care about philosophers’ intutions, but only whether our account fits the scientific evidence — if it does and the empirical evidence entails there is lots of MR, then philosophers need to get over their intuitions as they have so many times beofre in the face of scientific findings.
But, as I say, your worry is different. You are concerned because IF one allows the existence of all manner of entities like determinable properties (such as being a man or being an ion channel) THEN potentially the Dimensioned view allows as cases of realization and MR the relation between being a potasium ion channel and being an ion channel, and other particular kinds of ion channel and being an ion channel. But that looks dubious, you say.
I agree completely! But we need to be careful in noting WHAT is dubious. The mistake was in the antecendent of the conditional above where we let in determinable properties (like being an ion channel) as well as determinate properties (like being a potassium ion channel). I have argued against such determinable properties elsewhere (see my Nous 2005 paper with Brad Rives). IF you let in ontological junk, then of course the accounts of scientific realization and multiple realization give out wrong answers — but that was due to a prior mistake in what ontology one accepted and NOT a weakness in the accounts of realization and MR.
Did that make sense? The background point relates to my, and Ken’s, worries about sub-types: Once again, there is the worry that we are letting a kind of entity into our ontology that we have reasons to deny — now those reasons are empirically defeasible case-by-case but there are prima facie reasons to deny certain entities exist.
So the response to the first worry is that there is a problem, it relates to what we accepted into our ontology intially, and not to our accounts of realizaton and MR that work great with a cleaner ontology.
I respond to your other point in another post. best, Carl
Hi Eric, ok I like this question, too, since I have been thinking about identity theory again recently…
Your worry is that our accounts of realization and MR are “consistent” with the identity theory. And you give a nice scientific example, being a K ion channel could be multiply realized by the different properties and relations of different proteins, BUT being a K ion channel could still be identical to being a property certain functionally individuated properties. Something like that right?
Three things to say in response, which I don;t think have been emphasizied in past literature.
First, ANY account better be such that it allows identities of some kind. There is an obvious reason. Every entity is identical to itself, so true identities hold for every entity. Thus to demand a theory of realization and MR exclude all identities is obviously mistaken.
The latter point is not trivial because it relates to my second point. I suspect that the identity you outline is really just a self-identity: being a K ion channel just is being an individual with certain funcaitonally individuated properties of allowing certain kinds of ion, etc. So it is good if our account allows this kind of self-identity.
Third point: Ken and I are careful to emphasize that all realization and multiple realization claims are indexed to levels of properties. So we allow that a property P at level 7 could be identical to a property Q but also be multiply realized by properties R1-Rn and S1-Sn at level 6. So, of course, even for the same property we should allow the property is MR but that a substantive identity theory may hold of it — but with reagrd to properties at different levels.
So, again, I agree with the substance of your second worry, but once we qualify our claims I think our accounts of realization and MR should be, can be, and are, compatible with the point you make.
Hope that was clear, I bashed this out at one go with no corrections since I need to get back to writing this morning, best Carl
I have argued against such determinable properties elsewhere (see my
Nous 2005 paper with Brad Rives). IF you let in ontological junk, then
of course the accounts of scientific realization and multiple
realization give out wrong answers — but that was due to a prior
mistake in what ontology one accepted and NOT a weakness in the
accounts of realization and MR.
An interesting response. You can’t, however, be saying that it is a mistake to let K channels as well as v-gated K channels into your ontology. That would be nutty, no?
Regarding your first paragraph, you ask a good question. What justifies trichromat and dichromat that does not justify a further taxonomy? That’s a good question, but also, as you sense, not one I’m pursuing. The first task is to figure out what the actual practices are, then one can figure out what, if anything, justifies those actual practices. I’m only at stage one this month. Before I would pronounce judgment on the legitimacy, or lack thereof, of a scientific practice I would want to know more about the case than I do now. Working on a detailed, case by case basis is slow work.
Ok. Just to be clear, Carl and I are not denying that scientists soemtimes restrict their claims to subtypes. We know they do. What we are resisting is the move from the observation that the do at times hypothesize subtypes to what Richard seems to be after, namely, that any time there is an apparent case of MR, you can block it by introducing a subtype based on the different realizations. What I have been pressing is the “any time” part, where Carl with his “hardness-in-teeth” example has been pressing the “unacceptability in science” of introducing new subtypes based just on the different realizations.
Thanks for this response, Carl:
My second question about your paper was:
In K-channels,
does the ability to swap in functionally identical amino acid residues
into a K-channel imply multiple realizability is true? I think it does,
in your definition. But someone could accept this and still believe in
the identity theory of K-channels, no? Could someone say that the
K-channel is identical not to amino acid sequence, but the tertiary
structure of the sequence and the location and magnitude of certain
functionally important charges? Identity theory and multiple
realizability are both true, in this argument.
I
shouldn’t have put the words ‘functionally important’ in the
second-to-last sentence. Remove it, and it doesn’t change my claim. But
this is also what informed my above claim: one way to determine if two
K-channels use the same mechanism is to look at tertiary structure, the
locations of charges on amino acids (and the magnitude of the charge),
as well as certain structural features that are not multiply realized
(e.g., the disulfide bridge, which only forms between cysteine
residues).
Basically, I am exploiting the tight link between
structure and function in biophysics and arguing that an identity
theorist could identify a K-channel class with certain structures (by
structure I mean the stuff above: tertiary sequence etc).
So
biophysicists are interested in both realization issues (what residues
are where, and how is this different in different species) and
structural issues. Perhaps identity theory critics are wrong to focus
on realization as providing the relevant basis for an identity claim,
rather than structure.
This clarification makes me less sure that your responses work.
Implication
of my view: say there is convergent evolution and two organisms evolve
v-gated K channels. If the structure is different, they are different
k-channel types. If the structure is the same, they are identical types
that evolution converged upon.
This
implication is likely wrong, though. In biology we typically split
types using evolutionary relationships in addition to raw structural
considerations. Consider also mimicry–the poison red butterfly versus
the red butterfly that is mimicing the poison one. They are different
types despite (let’s assume) physical identity (I steal this example
from Sandra Mitchell’s JPhil paper from many years ago). They are
different types of coloration.
So, I may need to amend my view to include homologous structure, not structure simpliciter. Is such a view of identity conditions still an identity theory?
Note
I am surprised none of us hit upon this issue of homology when we were
discussions nonarbitrary ways to lump and split things in biology. That
is becoming the hallmark.
Stepping back, a couple of points.
First, this view of identity may be uniquely well-suited to enzymes and
channels and such, where the structure-function link is so tight.
Second,
the best argument against identity theory is externalism, twin
Earthians and such. All this realization stuff might be important for
general philosophy of biology, but for philosophy of mind there are
more direct arguments against identity theory. Of course, it would be
nice if all our theories converged. There is more than one line of
evidence for any scientific truth!
OK, this all seems reasonable.
Oh, in my recent post to Carl I mentioned one more (nonarbitrary) way that biologists like to subtype: based on phylogenetic relationships (e.g., you can have identical physical butterflies with different evolutionary histories: one a poison red one, another a mimic of that one). This might be relevant.
Regarding the move in italics. Yes, one could say that a protein is its tertiary structure. Bickle seems to want a move similar to this in reply to my 2007 Synthese article on the biochemistry of memory consolidation. And, in principle, it could work out as you say.
Once this logically possible move is on the table, then one can ask, “Well, do scientists actually identify proteins with their tertiary structure?” And, one has to be careful of assuming that they always do it this way versus only doing it this way some time.
FYI, the issue of homology as you seem to be thinking about it looms large in Bechtel and Mundale’s discussion of Brodmann’s work. So, they note that, say, macaque V1 and human V1 are the same in the sense that they are homologous. From there they implicitly suggest that macaque V1 and human V1 are unique realizations (of something).
I’m all for using established scientific results to constrain and guide one’s philosophy. There are philosophers out there that seem almost happy to be ignorant of relevant science (though they wouldn’t put it that way, they would say it isn’t relevant).
We can’t put too much stock in what scientists say about issues
that have strong philosophical currents. Scientists often make shitty
philosophers, advocating strange fads without empirical or reasonable
basis (e.g., Copenhagen interpretation/philosophical behaviorism).
I realize the above two paragraphs make it very hard to do good philosophy of the type you guys are doing. Ya’ have to steep yourself in the science, getting yourself to the point of competence where you can distinguish discussion points from actual results. To make it harder, scientists just don’t care that much about the words they use, as they are focused on the thing (in the world) they are trying to understand.
I should add one more type-making machine in biology: development. When cells share a common ancestral cell, a common fate during development, they are often grouped together.
The case of opsins is actually very interesting, as it seems much of the variability is generated by recombination during meiosis. This certainly makes it hard to come up with a finite number of types. There is an interesting article I just read over here. It is odd that it seems to only happen for the M and L cone opsins, not the S. They sometimes produce defects in vision, a fact that might have implications for this discussion. Blindness is multiply realizable, for instance! E.g., retinal (as with crazy messed up opsins), severing of the optic nerve, etc..
Check out Figure 1 in this article: Absorption spectra of the hybrid pigments responsible for anomalous color vision
SL Merbs, J Nathans – Science, 1992. The least you can say about this is that it’s a case that cries out for analysis by those interested in scientific treatments of MR.
Bickle is fond of saying things like, “Scientists don’t care about philosophical fantasies.” Right, but one implication of that is that scientists don’t much care about philosophical distinctions. So, my personal mantra is that I don’t need to be true to the scientists, I need to be true to the science. Of course, that’s easier said than done.
Ok. Here it is. May be hard to see and understand without the text, but the basic idea should jumps out: In the top row, the first two panels one finds hybrids pigments with apparently the same absorption spectrum as the green cone pigment.
Finally back from vacation, so I can actually look at papers on this.
Interesting paper, and they explicitly break up trichromats without normal vision into categories (subtypes?). From the first paragraph of the paper:
Also, some more general psychophysics has been done in a great paper from the same lab, one that is quite relevant for the mind-brain MR issue (link here):
Correlations between Deduced Protein Sequences and Psychophysically
Measured Spectral Sensitivities. L. T. Sharpe, A. Stockman, H. Jagle, H. Knau, G. Klausen, A. Reitner, and J. Nathans (1998) J. Neurosci.
18, 10053-10069
The abstract shows just how specific they get when discussing the relationship between amino acid sequence and psychophysical influences (e.g., ‘the alanine/serine
polymorphism at position 180 in the red pigment gene produces a
spectral [psychophysical] shift of ~2.7 nm’).
I’m not sure I agree these are the best kinds of examples for the study of MR. It may be better to stick with things that are better figured out (e.g., K-channel is a good one that you looked at in your paper, but even better would be much simpler, without subunits, whose structure is well-worked out, such as myoglobin, or the enzyme amylase). The better developed the science, the more detail you will have on hand to constrain the philosophy.
For those who don’t like MR, and have confirmation bias, you can always bring up ‘hereditary material’ as an instance of something that isn’t multiply realized.
Right, that’s where it bottoms out. You get to “anomalous trichromats” (which has that genus-species form), but nothing finer, at least to my knowledge. There are also lots that just fall into normal trichromats.
And, yes, there are some very precise measurements of the psychophysical effects of different amino acid sequences. I don’t deny that. In fact, I embrace that making my point against Richard. My point is that, in point of logic, they could introduce further genus-species subtypes based on distinct realizers, but they do not. So, insofar as Richard intends to describe actual practice, with his type/subtype idea, he seems not to be doing that.
No, it gets more specific. Look at the paper I cite. They are breaking it down to specific exons, to specific amino acids (as I cite). Is that not breaking down anomalous trichromats? If not, why not?
There is also this recent review article that breaks down anomalous trichromats into three different subdivisions.
At any rate, even if you were right about the linguistic practices of some scientists, that wouldn’t be a strong argument for the reasons I’ve said. Unless, perhaps, the scientists had obsessed about the question of the number of types, made it a topic of scientific study, and come to a consensus. This is why I brought up neuroanatomy, where they actually do argue about these things, but in the end they (most of them anyway) realize there isn’t a fact of the matter and try not to be too uptight about it.
But as you said, such issues may be on your plate, so I’ll stop badgering on about it. Clearly there are cases of multiple realizability in biology, but to pick cases where there are functional difference was your mistake I think. The strong cases are those in which you can substitute amino acid residues that have no notable phenotypic affects, like the opsins with the same absorption spectra that you showed. But those with different spectra, and different psychophysical consequences? Your argument is not convincing in such cases.
At any rate, myoglobin is one of the best understood proteins in the biological universe, so it would be relatively quick and nice study of the issue where most of the parameters, functions, structure, are all well known. It was the first protein to be cracked with x-ray crystallography so has received a ridiculous amount of study.
For that matter, just look at the paper you cited originally, which I quoted:
It looks like they have two categories here. You’re trying to pull some philosophical jive on me!
Ok. Fair enough. It breaks down into
deuteranomalous trichromat =deuteranomalous anomalous trichromat (DAT)
protanomalous trichromat =protanomalous anomalous trichromat (PAT)
tritanomalous trichromat =tritanomalous anomalous trichchromat (TAT)
You can get the type/subtype argumentgoing one more step. You got me.
Still, the logic of the case remainsthe same. The type DAT is a psychology type that can be identifiedthrough a Rayleigh test. (In the test, a subject gets a target colorlight, then tries to match it using dials that add greater or lesseramounts of primary color light.) Within the DAT type, there areindividuals with different numbers of green cone opsins and evenamong those with, say, three cone opsins, you can apparently findthose that differ in which three. Richard’s proposal predicts thatDAT will be further subtyped. It does not appear to be. Or, itcould be that Richard’s proposals recommendsthat DAT be further subtyped. Ok. Why follow this recommendation? Or, it could be Richard’s proposal is just an observation about alogically possible thing to say. Ok. I don’t want to challengethat, but so what?
And you are correct, that scientistswill classify individuals differently based on their genomes, butthat is orthogonal to the current issue, since it is not the DNA thatrealizes one’s color processing capacities, but the proteins.
I would also agree with you that thekind of case in which two amino acid sequences have the sameabsorption spectrum is easier to use to make a case for MR than isthe case where they do not have the same absorption spectrum. (But,Bickle cried “foul” even on this case.) But, I think I can dobetter than grab this low hanging fruit. But, you’ll have to waitfor the paper. I’m now thinking I have two more arguments that I’venot broached here.
You are also right that myoglobin hasthe virtue of being well studied, as does the sodium channel. (Hemoglobin doesn’t look so bad either.) But, I wanted an examplethat could relatively easily, and relatively uncontroversially, betied to something psychological. Bickle likes the biochemistry ofmemory consolidation, which has some nice features, but the biochemthere appears not to be as well worked out. The biochem of vision isa kind of compromise. The biology seems to be better than that ofmemory consolidation, although worse than myglobin or the sodiumchannel, but the psychological import of phototransduction is moreapparent than that of myoglobin and the sodium channel.
To the “So what?” It means the burden of proof is on you to show why we shouldn’t think there are different types (genetic differences are necessary, but not sufficient, for protein differences, so ultimately we ground them in genetic types). You made the strong, strange, claim that they wouldn’t subtype if they found psychological-genetic correlations, and said to do so would be philosophical jive. This put off my philosophy-bs working with genetics go about their business, as I’ve mentioned over and over. We split things up into subtypes if we found such correlations. That’s how the scientists work. That’s how I work.
But even if it weren’t how I worked, you still need a case independent of the sociology of my subculture.
And just because it is how we work, it doesn’t mean that is right. Maybe we split too freely. Most likely, there isn’t really a fact of the matter.
I agree that memory is murky. It is best to pick the simplest example that still has the features of interest. If the structure of the opsins is not worked out (x-ray crystallography) that might be a relevant thing to consider. Myoglobin was the first structure to be revealed via x-ray cryst.. Sodium channels would be even worse than the potassium, more complicated, less x-ray cryst done.
At any rate, I am not convinced at all by your argument, and think you should marinate yourself in the subculture of genetically-sensitive neuroscience (C elegans and mice).
I cut off a sentence, and just generally mangled it. I said:
I should have said:
This set off my philosophical-bs detecter, because it is not how people working in molecular neuroscience go about their business.
To the “So
what?” It means the burden of proof is on you to show why we
shouldn’t think there are different types
But, I’m not trying to press a
normative claim about what neuroscientists or psychologists should
be doing. At this point, I’m
just describing what they are doing. Maybe my claims are
disappointing to you, since they are not up to what you think I
should be doing. Is that it?
You
made the strong, strange, claim that they wouldn’t subtype if they
found psychological-genetic correlations
Not
exactly. I say that they would subtype if they found
psychological-realizer correlations. I say that, given the
opportunity to create psychological subtypes (on the model of
genus-species) on the basis of differences in realizers,
neuroscientists and psychologists do not so subtype. I am proposing
that Richard is making an incorrect descriptive account of what
happens in this area. (Maybe Richard is right about what they should
be doing. I’m not that far along, but then again, he’s not given an
argument for this.) I concede that neuroscientists and psychologists
care about genotypes, but to divide by genotypes is not to divide by
realizers. The DNA doesn’t realize the visual apparatus, the
proteins do.
So,
just tell me, what are the dozens of psychological subtypes of deuteranomalous
trichromat based on the different amino acid sequences? There are
lots of instances of this. Just read the literature and tell me how
it goes. You know how to refute my view. Find some psychological
level types that match the different neuroscientific realizer types.
I already pointed out many examples of what you said they wouldn’t do. Read the papers. I never said there would be dozens already done, but there are some in the second paper. I didn’t even assume any subtyping would already be done. I just know this kind of work well enough to know what the obvious experiments (and reactions) would be, and I turned out to be right. It’s not fast work, as evidenced by the time between the original paper and the second one I cited. I have made my case in all these posts and provided a bunch of objections and potential objections to your argument.
Stick to functionally identical variants, and you’ll be fine. Once you enter functionally different variants, you are not. Addressing the normative dimension is crucial, as it is the crux of the matter (that is, unless as I already stated the science was focused on this question, as it has been in anatomy, or periodic table of the elements, or whatever, then perhaps you would have a little more footing upon which to base a methodologically naturalistic argument).
Actually, it is not clear that you have
any examples of what I say
wouldn’t happen. Let me try to be clearer.
Richard’s
claim, construed descriptively, is that given a type and a potential
diversity of realizers, scientists will introduce subtypes. I reply,
“Not always.” You seem to agree and think that what Richard says
is too strong, so that instead it should be something like given a
type with diversity in it and
a diversity of realizers, scientists will introduce subtypes. My
reply again is, “Not always.”
I’ve
tried to describe an area of science, human color vision, where this
putative iterative process stops. I’ve tried to give an example in
which there is a psychological type with diversity in it and a
diversity of potential realizers, but where scientists will not
introduce subtypes. For the sake of concreteness, I proposed that,
in fact, the process bottoms out at “anomalous trichromats.”
That is, incorrect, as you noted. It clearly goes on (at least) one
more step. What I was wrong about is that “anomalous trichromats”
is the current finest subtype.
Now,
you apparently think that this shows something that I say would not
happen. But, of course, my claim is not that scientists don’t create
subtypes at all. (They clearly do. As noted there is pain, human
pain, octopus pain. I mentioned dichromats and trichromats in
another post months ago. You mentioned memory, long term memory and
short term memory. So, this is familiar.) There is no
misunderstanding here.
So,
you might think my claim is that scientists don’t ever
create subtypes once they have discovered a diversity in
potential realizers. You might
think that my view is, to put it in a new way, that the discovery of
a potential diversity of realizers is never
an agent of scientific change of the sort that introduces subtypes
into types. And, you might think that the deuteranomalous trichromat
case provides an example where they do create subtypes in
response to a discovery of a diversity in potential realizers.
But, here is where the lack of clarity in the case comes in. It is
clear that a new subtype appeared, deuteranomalous trichromat, but it
is not clear why it appeared. In particular, it is not clear that
this subtype appeared in response to a discovery of a
diversity in potential realizers,
rather than to meet some theoretical needs in psychology. So, the
reason that “deuteranomalous trichromat” was introduced is not
100% clear. That’s why it is not clear that you have any
examples of what I say wouldn’t happen.
But,
that aside, it seems to me that not much rides on me making the
strong claim that scientists don’t ever
create subtypes once they have discovered a diversity in
potential realizers. I could
probably have an open mind about how often this works this way.
Only one thing I’d clarify–I don’t resist your philosophical claims for all realizers (e.g., opsins with the exact same absorption spectra), only for the case of those realizers with diverse functions. I think multiple realizability exists. The problem with this argument about subtyping is that ultimately I think it will not have legs, compared to Carl’s second argument. It will come down to a war of intuitions about how fine-grained our categorizations of the world should be.
Note that narrowly interpreting Richard’s view (to the extent that his view was well-formed, which really isn’t apparent from his post) as something about what scientists actually have done, is uncharitable. The question is how the world should be carved up based on the evidence, regardless of the (often contradictory) classification schemes of neuroscientists. What is the fact of the matter? If that doesn’t have an answer, then what’s the point?
Finally, on a somewhat different topic. Carl’s response to my argument about the promiscuity of the ‘Dimensioned’ view of realization suggests the view has some strange consequences, consequences that scientists clearly don’t accept (e.g., that K-channel and v-gated K-channel are not both legitimate categories). Perhaps this dimensioned view needs more constraints to do justice to our categories, not vice-versa (note I am using your arguments about scientific categorization practices against you here, not necessarily advocating them in my own voice).
As it is, the dimensioned view seems to provide little more constraint than the supervenience relation between sets of properties. Indeed, they are so close I wonder if you’ve published about the relationship between supervenience and you/Carl’s view of realization? Roughly your view seems to be:
G ‘dimensionedly realizes F’ =def G is sufficient for F, but F is not sufficient for G.
F, G are lists of properties. This is very close to supervenience. It seems different from supervenience in a subtle way. It implies that if you change F, so that something now has property FF (different from F), that G is no longer present. Also, as with supervenience, it seems DR implies that identical G implies identical F.
Would your view imply that changing F-properties implies a difference in G-properties? It’s not clear, as the DR view would probably need to be generalized to even answer that question.
So getting to your positive story, I think it is incomplete. More constraints need to be added, some guts added to the DR exoskeleton to differentiate it more clearly from supervenience, and to conform to K-channels.
Right. That was what I was trying to get at in the second paragraph of my last post when I wrote,
You’ve said this kind of thing a couple of times and I’ve ignored it, but I just don’t see that a descriptive project of scientific categorizations has to get bogged down in a normative dispute about how fine-grained our categories should be. I keep saying that I’m working on a descriptive project, right? And you disparage that descriptive. But, you insist that I have to do some normative project because that is the “crucial” issue and then you disparage the normative project as a boring dispute over which there is no fact of the matter.
Well, I’m not telling Richard what he has to argue or what the crucial issues are. All I am saying is that, construed in one way, it appears to have these consequences. That allow for other interpretations. And in fact, in an earlier post, I hinted at three other interpretations:
Ok. My question is not the question. I can live with that. But, to repeat the point I made above in this post. You seem to be of two minds about the normative of how we should taxonomize. Sometimes you say things like it’s the issue and at other times you suggest that there may be no point.
Yes. I think that basing arguments that these subtypes don’t exist on contingent things is wrong-headed, and you need a normative criterion. However, I am skeptical that such a normative criterion exists. That is why I don’t like this whole approach, and prefer Carl’s second argument against Richard.
But if you must persue it, to find out how (one group of) scientists went about describing their discoveries, the follow up paper I cited looks to be a good one to start on as a window into molecular neuroscience. You would need to address the specific psychophysical correlations they discover, and justify the claim that the scientists aren’t referring to different subtypes of anomalous trichromacy (regardless, of course, of the language they use to categorize the different phenomena; you still would need to determine if they were talking about different categories even if they didn’t use those words). Since the scientists are still alive, you could probably ask them. My bet is that they would be like, “Huh? Who cares? All these things are just labels, and we are interested in the phenomenon itself. Call them what you want. We need to get back to work.”
I’m in the middle of Shapiro’s ‘Multiple Realizations’ (JPhil 2000). It is very good, many of the issues that we have discussed are brought up in there. He seems to be right a lot.
JPhil should require abstracts. That journal is so freaking old school.
Yes, Larry is great and usually right, except when he disagrees with us. Alas, there is an emerging literature of Shapiro-Polger versus Gillett-Aizawa. At some point, you have to get past fighting each other and get on to the cases.
Making matters worse for me is that Larry has a paper, “Testing MR” or some such, forthcoming in Philosophy of Science, where it looks like he is arguing that you cannot get MR in cases where there is difference in the higher level property. So, he’s exactly on this issue. But, I’ve not yet gotten my head around the argument.
Hey everyone, super interesting discussion here!
Sorry I haven’t been around, I was busy (successfully) defending my dissertation!!!
Congrats, Richard!
Congrats Richard!