Quick and dirty, short and sweet: Yes. Essentially every single known psychological
property is multiply realizable by many, many distinct amino
acid sequences.
Are all psychological properties multiply realizable?
24 Comments
Comments are closed.
This seems to be the sense of multiple realization also used by Derk Pereboom in his recent paper, “Robust Nonreductive Materialism”.
Ok, but is this an interesting kind of multiple realization?
From what I remember, Larry Shapiro (in his 2004 book, Mind Incarnate) has an argument to the effect that simply replacing components of a system with other components that contribute the same causal property to the whole should not be counted as multiple realization in the sense that traditionally interests philosophers of mind.
What do you think about Shapiro’s argument?
Thanks for the reference to Pereboom. I will check it out.
Whether this is an interesting kind of multiple realization is a good question. Maybe not. Maybe the interesting question is not “Is X multiply realizable?” but rather “Is X multiply realizable by things at level Y”. I’ll leave that for now….
The reply to Shapiro is that difference amino acid sequences do not contribute the same causal properties to the whole, but only similiar causal properties. Different sequences differ in mass, charge, shape, binding coefficients, reaction rates, etc.
I think Shapiro is right about the science, wrong about the philosophy. Different amino acid sequences often have the same relevant causal properties. Example: for maintaining the shape of an ion channel across the cell membrane, one of the more important properties is hydrophobicity (a hydrophobic string of residues is embedded in the membrane, while a hydrophilic string of residues hangs outside the membrane). There are all sorts of residue substitution experiments in which mutant ion channels are expressed in frog oocytes to determine whether the channels are functionally affected (sometimes they are, but often they are not (and predictably not)). If you change from one hydrophobic to a different hydrophobic residue (and this is in the middle of a chain of hydrophobic residues, and you don’t drastically change the charge/mass), you will not affect the function of an ion channel in which the residue occurs. Here is a link to an example article.
While it could later be discovered that there are subtle effects, right now it seems that some aa substitutions don’t influence the function of some ion channels (the most relevant aa chains in brains).
At any rate, I think Kenneth takes the wrong response to Shapiro. It seems clear that, as summarized, Shapiro is just wrong that Kenneth doesn’t get at the meaning of multiple realizability “that traditionally interests philosophers of mind” (‘traditionally’ meaning ‘since around 1968’). Kenneth’s is clearly the sense originally used by philosophers like Putnam (and illicitly used to jump to functionalism as opposed to identity theory). The philosophers’ claim, in support of functionalism, was that you can substitute things made of different stuff and have the same mental properties. Indeed, one of the dog-show examples is substituting silicon for carbon.
If Shapiro wants to, instead, say that such dog-show cases don’t help us decide among different versions of functionalism (and identity theory, which is consistent with some forms of functionalism), then he should say that, and point out how interesting it is given the tradition.
Is there a way to get a ‘preview’ function on these posts?
Eric,
Thanks for this comment.
I think that in proteins one does not need to maintain exactly the same causally relevant properties, e.g. shape, for, say, the ion channel, but only “close enough.”
Variants in amino acid sequences come in different types. There are
Variant sequences that vary in causally relevant properties and retain functionality
Variant sequences that vary in causally relevant properties and do not retain functionality.
Variant sequences that do not vary in causally relevant properties and retain functionality.
It is the first type that give rise to multiple realizability. There can well be cases of the second and third type, but that’s ok as long as there are cases of the first type there will be multiple realizability.
In these kinds of experiments to which you allude, there is only an exploration of functionality. I think it is taken for granted that there are often small differences in causally relevant properties. But, I could, of course, be wrong. I’ve not surveyed the whole of the literature.
Note that one can change the causally relevant properties of functional domains by making changes in amino acid sequences in those domains, but also by making changes in amino acid sequences outside those regions. This gives rise to more “opportunities” for different realizations.
There are, of course, other things to address in Shapiro’s view, but I do like to blog short replies.
Eric,
In the abstract to which you refer, there is the observation that:
Mutation of acidic residues (D540, E544) in the S4-S5 linker of HERG
channels to neutral (Ala) or basic (Lys) residues accelerated the rate
of channel deactivation. Most mutations greatly accelerated the rate of
activation. However, E544K HERG channels activated more slowly than
wild-type HERG channels.
I think this makes the point I am driving at with regard to Shapiro more concretely. There are amino acid variations that are causally relevant, e.g. influence the rate of activation or deactivation of a channel, but which retain functionality, e.g. the channels still activate and deactivate. Here I am presupposing that the channel can retain a given functionality, with variations in the rate of that functin. That, of course, presupposes that this is an acceptable way to individuate function.
There is no ‘preview’ function. If you are writing a long post and you wish to review it, perhaps it would help to write it in a separate file before you publish it.
I take those changes to be functionally relevant causal properties (activation/inactivation rates are very important). I guess I don’t understand what you mean by ‘function’ if you say there can be changes in ‘causally relevant properties’ without a change in ‘function’. I was using ‘function’ as it is used in the in biophysics of ion channels literature: changes in the electrical phenotype of the channel are functional changes. To have ‘implementation’ changes w/o such changes you need different aa sequences without electrical changes.
My point in citing the above paper was just to point to the rather large literature. Most papers on such substitution experiments will have some functional change (otherwise it won’t get published: when there is no change it isn’t very interesting), but there are papers where they observe lack of changes, such as this one (the Leu54Ile change produces no change in the electrical phenotype).
You could get what you want easily: instead of one sodium channel with activation/inactivation dynamics you can posit two sodium channels, one which follows typical activation dynamics, the other which follows inactivation dynamics, but generate the same electrical phenotypes. No reason this can’t happen in principle.
This focus on amino acids seems like a red herring.
Take an amino acid sequence with activation/inactivation rate X and an amino acid sequences with activation/inactivation rate Y. I would say that they both have the function of being, in this case, ion channels.
To get multiple realization of a property I’m supposing you have to have two instances of the property produced by two causally distinct mechanisms. If your view is that every difference in causally relevant properties is a difference in function, then this won’t work. If you think that an amino acid sequence with activation/inactivation rate X has one function, where an amino acid sequence with activation/inactivation rate Y has another function, and there is no notion of function like the one I want, then this won’t work. This seems to me what you are urging in your first paragraph in your last post.
Note that I am not talking about the functions as might be identified in biophysics being multiply realizable. I’m talking about ‘essentially every single known psychological property”. They are the ones I’m arguing are multiply realizable.
The focus on amino acid sequences seems to me good because it is a very general feature of life as we know it. It gives me a very simple way to make a case for multiple realizability of essentially all psychological properties. You can basically get a whole philosophy paper boiled down to one sentence. (Incidently, Gualtiero says I should mention that I have a paper forthcoming in Synthese on this topic. The proof is available on my web page.)
That paper is at:
https://www.centenary.edu/attachments/philosophy/aizawa/publications/aizawa2006.pdf
Your paper focuses on certain types of memory, for which a molecular genetic approach is probably most appropriate. An extension to all psychological states, which is your claim, is not warranted from analysis of such a model system. E.g., there is nothing about representations there.
Amino acid sequences are very important, but so are lipids (their shape/surface area determine membrane capacitance), carbohydrates (glycoproteins), gases (nitrous oxide), ions/ion concentrations. It seems very odd to focus on one ingredient in the functional story of how brains work. It’s like saying that car engines are multiply realizable because spark plugs are multiply realizable. Is this really true? It isn’t obvious at all, so your arguments are not convincing.
Also, why not focus on something that is already well worked out, like oxidative phosphorylation or something? The mechanisms of memory are in a state of flux. I think the best philosophy of science works where the science is already well worked out (look at poor Bechtel’s prediction that the Na/K pump didn’t exist, that it was likely a ‘localization error’ in an otherwise excellent paper on functional analyses :)).
Plus, by explaining synaptic plasticity in vitro, have they really explained anything about memory as a psychological (representational) phenomenon? There is scant evidence linking these lower-level studies to psychological studies of memory.
So, while what you say may well be true (I’d need to read your paper more closely), i.e., certain events at the molecular level (phosphorylation) are multiply realizable, it seems to not directly address issues in philosophy of mind. This is fine with me: we need more good philosophy of biology, application to simple model systems, rather than speculations about human brains.
I wonder, Ken, what you think rides on this. Unless I hear a little more specificiation of what does and doesn’t count as multiple realization, for all I know, all non-psychological are multiply realizable as well.
Eric,
Thanks for your interest in this topic.
As you have gathered, the idea in the paper is to generalize from one model system. Amino acids are important components in lots of biological structures. Your analogy with the spark plugs is exactly what I have in mind. This one (important) component is multiply realizable so the whole is. I don’t think it is as extreme as the spark plug case, but you have the right idea. I would say that, roughly speaking, proteins are more prevalent in the body than spark plugs are in cars. I guess I don’t see a problem here, but I’d be interested in a counterargument.
I guess I could focus on the multiple realizability of oxidative phosphorylation, but then I would need to have an argument connecting that to psychological properties. I’d need something tying the metabolic process of oxidative phosphorylation to thinking. It might work, but it’s not obvious to me how.
I don’t see that I really need for this biochemistry of memory consolidation science to work out that much in detail. The basic idea, which has been the topic of this post, is really that any psychological property we know of now will, in biological systems like the ones we know now, involve multiple proteins and each protein can correspond to distinct amino acid sequences.
But, my point is not in the first instance that, say, phosphorylation is multiply realizable, but that psychological properties are multiply realizable. This seems like it might address some issues in the philosophy of mind. I don’t see this as all that speculative. What of it seems speculative to you?
Pete,
I see four things sort of riding on this.
First, this shows, I think, that there is no identifying psychological properties with biochemical properties. This is in opposition to Bickle, “There is a ‘physical-chemical state,’ the cAMP-PKA-CREB molecular pathway, that uniquely realizes memory consolidation across biological classes, from insects to gastropods to mammals” (Bickle, 2003, p. 148).
Second, one wants to know about inter-level relations in science. Carl Gillett, for one, has a theory that includes a theory of a realization relation and a theory of multiple realization. Gillett, it seems to me, provides a nice philosophy of science kind of account of the relation at work in the biochemistry case.
Third, it seems to me that Shapiro’s (2004) theory of realization can be informed by this example. His theory appears to apply to the case, but not to warrant the conclusion that Bickle wants.
Fourth, Bechtel & Mundale, (1999), Shapiro, (2004), and Polger, (2003) suggest that a closer look at (neuro)science will reveal that the hypothesis of multiple realizability of psychological states is less plausible than previously suspected. This example runs counter to that trend.
There is probably a lot more riding on this, but these are the ones I see now.
The theory of realization being invoked here is from Gillett, (2000), his JP paper. The theory of multiple realization is given in a paper Carl and I hope to wrap up sometime in September. I can send you a draft if you like. Or I guess I could post it (I’d want to double check with Carl.)
Yes, it could be that even all non-psychological properties would be multiply realizable. What of it?
Thanks, Ken. That helps quite a bit. I’d be interested too in see what you and Carl are up to.
Re non-psychological properties, if they are all multirealizable too, then all properties are multirealizable and any partciular claims re multirealizability seem kind of vacuous. The classic “A and B are made of diffferent kinds of stuff therefore whatever properties A and B have in common are multirealizable” inference strikes me as pretty underwhelming. however, in the hands of some people, the next conclusion drawn is that if the realizing properties are physical, then the realized property is nonphysical. That strikes me as insane, to put it mildly.
Why would this be insane? Well, take a property like mass. A and B can have the same mass, but be made of totally different stuff. therefore mass is multiply realizable? boring. therefore mass is nonphysical? insane.
So, I guess my challenge is, what is multiple realizablity such that it is neither boring nor insane? It’s been a while since I’ve read Shapiro. Is it fair to say his answer would be “nothing”? Also, Ken, what’s your answer?
One thing you might have noticed was not on my list of four
things was “implication for reductionism”.
I’m not yet sure what this kind of multiple realizability holds for
reductionism.
entertaining now. Yes, the question, “Is
X multiply realizable?” or the thesis that ‘X is multiply realizable” is
boring, since for most all high-level scientific properties the answer is
yes. A less boring question is, “Is
there a level at which X is not multiply realized?” If the answer to this were yes, then one
would have something that looks like it might be a level suitable for a type-type
Nagelian reduction. On this reading,
challenging an old-fashioned type-type identity theorist for property P would
require more than just showing that P is multiply realizable. It would require showing that P is multiply
realizable at all levels. So, perhaps P
is not uniquely realized by anything at the amino acid level. But, maybe P is uniquely realized by
something at the neuronal level. If so,
then maybe P is reducible to something at the neuronal level.
(Incidently, this note goes back to a hint in my first reply
to Gualtiero.)
(Your case of MR of mass is interesting. I’ll have to think a bit more about it.)
Some other physical, but multiply realizable, properties (some are stolen from JJC Smart and Polger, I think):
-energy
-brittleness
-phase (liquid/gas/solid)
-velocity
-volume
-temperature
-viscosity
It is strange that these examples have yet to become the quick and ready responses to blithe claims for functionalism based on multiple realizability. In theoretical neuroscience, the analogy with temperature (and other macrostate variables from statistical mechanics) has been pretty common since Hopfield.
Eric,
These are good examples and merit some attention. NB: My post did not begin with functional properties, just properties. This is part of the reason why my post to Pete Mandik regarding implications does not mention functionalism, among a number of other important themes surrounding MR.
Hi,
I am sorry to interrupt, but briefly about this:
“Fourth, Bechtel & Mundale, (1999)… neuroscience will reveal that the hypothesis of multiple realizability of psychological states is less plausible than previously suspected.”
Well, at least S. Kim (not Jaegwon:)!) has argued quite convincingly that there is something wrong with the B&M- paper. Kim, if I remember correctly, argued that the case B&M make is only about homological traits. Homologies are characters shared between the species that are inherited from their common ancestors. Homologies provide neither evidence for nor evidence against the multiple realizability thesis.
Kim himself, as far as I remember, demanded the “homoplasies” instead. A character shared between species is a homoplasy if it is not inherited from their common ancestor. He argued that if brains would have evolved several times in different lineages and still have the same physical-cum-structural properties, then there would be evidence against the multiple realization.
This was just an independent note, and it is probably not going to help a bit…
There is a lot going on in the B&M paper. Here is one defense I think they might try against
S. Kim. They are only claiming that
psychological properties are not multiply realized,
not that psychological properties are not multiply realizable. If a given psychological
property P is only realized by homologous structures (and all homologous
structures count as unique realizations), then P is uniquely realized.
(2003).
textual evidence that it is (cf. B&M, p. 177), then S. Kim is missing their
point.
understanding of what I think B&M are up to.
psychological property is multiply realizable by distinct amino acid sequences
is to claim that all the amino acid sequences that realize it count as one
realization because they are homologous.]
Yes, a good point. If B&M are truly arguing only that psychological properties are multiple realized, then it is ok.
However, is their argument then strong enough against Putnam`s original argument? I may remember uncorrectly, but it was about multiple “realizability”, not realization… But to be honest, I may be wrong here.
BTW; couple of years ago I was trying to understand something about the systematicity- debate. Then I found your article (from Mind and Language), it was extremely helpful. Thank you for that.
I agree that one should wonder whether challenging actual multiple realization does that much to challenge multiple realizability.
Thanks for your comments regarding my systematicity paper. I also wrote a book on it, but there is a short article that gets at the main point:
Cognitive Architecture: The
Structure of Cognitive Representations.
(2003). Warfield, T., and Stich, S., Blackwell=s Companion to the Mind. pp. 172-189.
There’s a book you might have on your shelf for other reasons.
I was so sorry, since I did not know about your book when I was… fighting with the systematicity-issue. And I haven`t seen your book here…
However, I always thought that your compositionality argument was really good one. It was really interesting, especially the Turing machine- example.
By the way, do you remember Matthew`s paper in the same journal? He critized Fodor by saying that Fodor is making very suspicious metaphysical speculations and so on. What did you think of that claim?
I´d love to write a bit longer, but have to run…
Don’t worry at all about not seeing the book. Few people have. It came out in 2003, which I think is long past the time when the issue was of much concern to philosophers.
I’m not certain what part of the Matthews paper you are referring to, but I’d be more than happy to follow up on your question.
In a hurry and with “finglish”:
Yes, I know many people think that the systematicity issue is “passè”. However, I personally think that the questions are still really important.
I do not have the Matthews- paper in front of me, and if I do not remember it correctly, thousands apologies to mr. Matthews:). However, if I remember correctly he wrote about the C- condition for systematicity (in Fodor and McLaughlin, 1990 I guess) and said something like that it is just a metaphysical speculation to claim that all the constituents of representations in M refer or express the constituents of propositions, and how the principle of co-existence is based on metaphysical speculation etc.
I thought that the argument he offered was… cheap:). Is it, even in principle, possible to formulate such an account of systematicity that would not be committed to “referring the elements of propositions” and the principle of co-existence?
Secondly, I have always thought that the main problem with the systematicity is the lack of empirical data, not the metaphysical speculations. However, there is a difference between empirical data and empirical data;-). You know probably the paper written by Van Gelder and Niklasson in 1994 (?). The paper in which they argued that the results from Wason-tasks can be seen as counter-examples to systematicity? But did the Wason task- experiment really measure the systematicity? The inferences used in Wason tasks did not include _same_ constituents, and it is a bit hard to see, was they really about systematicity.