FAPP have replied to Ned Block’s and Phillip Kitcher’s critique in the Boston Review, showing that the interpretation I gave before is the right one: they really do think that because we cannot say without ambiguity, a priori, what it is that natural selection is selecting, and therefore there is no theory of natural selection. Moreover, Elliot Sober confronts Fodor in a Bloggingheads.tv discussion, and in effect asks, “Why didn’t you read my book from 1984 on this?”
Home » More on the Fodor and Piatelli-Palmerini thing
Um, the link is to Basl’s Facebook account. 🙂
Thanks. Trick of the brain. I removed it.
Typo corrected, thanks to Jim Foley’s eagle eye.
“…because we cannot say without ambiguity, a priori, what it is that natural selection is selecting” …”
But we can! Long-term reproductive success. Can we reinstate the theory now, then?
I think I’m on Fodor’s side here; maybe this would be a good place to hear some opposition!
I am particularly convinced by the analogy given toward the end of the bloggingheads video: give me masses, positions, and velocities, and I’ll give you, via Newtonian mechanics (and nothing more), the future evolution of the system described. Give a biologist enough information about phenotypes and ecological variables, and she’ll give you the future evolution of those phenotypes, but not via natural selection alone. Rather, the biologist will use detailed knowledge about the particular situation, and those details are not like the details in the Newtonian mechanics example, where the details are kind of like parameters or values of variables that the theory specifically has places for (as Fodor puts it, the theorist is much smarter than the theory, which is not true when it comes to Newtonian mechanics).
Further, even if we come up with mathematical models, these are still not going to be general enough to count as anything like laws. They may be really good explanations for particular cases (or even families of cases), but they will be like using game theory to augment the historical explanation of events in, say, political history: they will not amount to a general theory of history.
It seems right to note that natural selection is not the only game in town when it comes to modern evolutionary theory, but insofar as it is a part of it, I’m curious about why this part of Fodor’s argument doesn’t work.
I think that is true, but this turns on the nature of explanation and theory in science. If you adopt Fodor’s approach, then nothing in science is a theory unless it is a general physics. Even the usual physical explanations fail because they do not tell you, a priori, where Pluto will be or whether the Oort cloud will deliver a planet buster on Tuesday next geological age. The constraint here is too great.
But as it happens I agree with some of his argument, as I have previously argued on this blog – the principle of natural selection (PNS) is not an explanation, but an explanation sketch, as Robert Brandon had previously argued also. Part of the trouble with Fodor is that on the one hand his rhetoric is overblown, and on the other he is reinventing the last thirty years of the philosophy of biology, starting, as philbio did, with Ernst Mayr’s characterisation of laws in biology. Sober was very restrained.
I personally think that there is no selection “for”, as Sober has defined it; there is only selection. In fact, I think that follows from Sober’s own discussion in The Nature of Selection. That the PNS cannot tell you what it is that has a given fitness in a given case merely means that the PNS is a formal modelling process that needs to be interpreted; but such models hardly mean that “Darwinism”, whateverthefuck that might be, is false. It means that any given application of the PNS must be interpreted in extra-Darwinian ways. Which is exactly what Darwin himself, and pretty well all those since, have said.
To argue that there have to be Darwinons or some ontological physical object that the PNS requires is silly. It is enough if you get things that satisfy the constraints of any selective process, and two candidates are replicators (Dawkins and Hull) and reproducers (Griesemer, Wimsatt, Godfrey-Smith). I personally go for the latter. These are substrate neutral classes of things.
The fascination, nay, obsession, some philosophers have with laws comes at a time when laws are roundly ignored by modern physics, in favour of general models. Models are context sensitive, multidimensional, and contingent to initial and boundary conditions, and apply equally well to any physical case where the constraints are satisfied. Consider weather systems as special cases of thermodynamic systems. They are not entirely predictable, but they are generally explained in terms of dissipative structures.
So I half agree with Fodor but disagree with how he deals with these considerations (and in any case he should get to know the literature first, which evidence suggests he hasn’t if Sober has to explain the selection for/of distinction). But I can’t read the book right now and own’t for another two months due to pressures of work.
Fodor’s logic is impeccable and revolutionary. It will change not only the way we do biology, but every other scientific field as well.
For instance, Fodor’s logic proves that Newtonian gravity “doesn’t work”. Gravity (supposedly) exerts an attractive force between two masses, proportional to each mass and inversely proportional to the distance between them. But gravity cannot tell the difference between (1) things with mass and (2) things lacking mass but physically coupled to things with mass. It will pull on them both just the same. As neo-Newtonians understand it, gravity is not able to distinguish the causes of acceleration from their local confounds.
As one way to see this (not by any means the only way) — the gluons within atomic nuclei are massless, yet they get accelerated by gravity by consequence of the protons and neutrons they are correlated with. Thus the free-rider problem — massless gluons literally ride free with protons. Prima facie, free riding is a counterexample to gravity. As far as I can see this objection to gravity is decisive.
I don’t think this analogy works. Gravity never could “pull” on things without mass, whereas free riders in biology could be operated upon. It would be confused to ask whether gravity is pulling on the basketball (which has mass), or the orangeness of the basketball (assuming orangeness is a massless abstract property). Orangeness never could play a role in Newtonian mechanics, whereas mass clearly does. But whether something is a free rider or not is a substantive question in evolutionary biology: it does not seem to be confused to ask whether it was trait 1 or trait 2 that was selected for.
Okay, here’s what I take to be your argument:
You’re saying that in the case you give, you have something free-riding: massless gluons. But it would be absurd to think that the free-riding of massless gluons shows that gravity doesn’t work. Am I wrong in this characterization?
I’m saying that’s a non-starter, because massless things never could be acted upon by gravity. This is not controversial, given that NM is explicitly about only those things that have mass. They are not free-riders in the way that traits sometimes are free-riders, and sometimes aren’t. The only way your case would be analogous would be if we had to run experiments to find out if gravity were acting on the mass of the particle, or the massless gluons (or some other property). We don’t: gravity only acts on mass. If it doesn’t have mass, it can’t be a free-rider. NS only acts on traits: if it’s not a trait, it’s (obviously) not a free-rider. But if it is a trait, it might be a free-rider. There’s just no analogy with gravity.
In your second example, it’s more complicated, because you invoke another force, with another kind of property acted upon by that second force. Of course this happens in biology: nobody now thinks that NS is the only game in town.
As to the counterexample comment, I was only pointing out that the second scenario you describe would be a counterexample to the claim that NM is the only game in town (thus it would not be absurd to think it was a counterexample to NM, and in fact, there was a time in history when it was a counterexample, because we did think NM was the only game in town). In biology, I think it’s clear that genetic drift and several other factors are involved in evolution. But that doesn’t mean that we can’t try to make sense of NS on its own, just like physicists try to make sense of gravity on its own, while acknowledging that particles can be acted upon by multiple forces simultaneously. I think it’s actually rather important, because if we get it right, we can figure out precisely what contributions the various forces (at least in physics) contribute.
As John pointed out above, it may be that Fodor (and many others) have the wrong conception about NS and “selection for.” But your analogies here don’t work.
Corey — You have characterized my argument correctly. However, you are assuming, erroneously, that we know a priori that gluons are massless. This is of course the current theoretical understanding, but we don’t know this — God hasn’t told us yet. Experimentally, the current evidence is that they are less than 20 milli-amu or somesuch. You state:
The only way your case would be analogous would be if we had to run experiments to find out if gravity were acting on the mass of the particle
Well yes, in fact, we have had to run experiments to see if gravity were acting on gluons. And such experiments are still being run. There are some variations of current theories that predict massive gluons.
Your objection is like me saying “of course non-adaptive traits are free-riders — non-adaptive things never could be acted upon by selection. This is non controversial, given that NS is explicitly about only things that have fitness differentials. … selection only acts on fitness differentials. If there is no fitness differential, it can’t be a free-rider. Etc.”
Whether a certain trait is non-adaptive or not, and consequently whether it is a selection free-rider or not, is the very point. Analogously, whether a certain particle is massless or not, and consequently a gravity free-rider or not, is the very point.
Thanks for continuing this discussion; this is what counts (at least some of the time) as research in philosophy!
I thought, from your original characterization: “Thus the free-rider problem — massless gluons literally ride free with protons” that it was being assumed that gluons don’t have mass. I take the following to be uncontroversially true: if gluons have mass, they will be acted upon by gravity. If they don’t, they won’t. We do have to determine experimentally which of these ways the world works.
Here’s what I take to be a free-rider in biology: a free-rider is a trait that could have been acted upon by selection, but in fact, was not. Rather, what was actually acted upon is a trait coextensive with the free-rider. Sometimes it’s the insulating properties of the brown fur that gets selected, and in a different context, it’s the color of the brown fur that gets selected. If a trait couldn’t possibly have been selected for (such as, I don’t know, whether a particular mammal has a number of individual hairs that’s divisible by 13), then it’s not even a candidate for being a free-rider.
Now, if gluons have mass, they aren’t free-riders, because gravity acts on things with mass, and if its being acted upon, then clearly it’s not a free-rider.
If gluons don’t have mass, they aren’t free-riders, because they couldn’t be free-riders. They are not something that gravity could have acted upon, but in fact didn’t. Gravity couldn’t have acted on gluons at all if they have no mass.
Thus, a gluon, massless or massful, can’t be a free-rider on gravity.
But consider insulating properties of fur: if selection did act on it, then it’s not a free-rider. If selection didn’t act on it, then it is a free-rider.
That’s why I think your analogy fails. Gravity free-riders do not make sense to me, precisely because there’s only one kind of thing gravity can act upon: namely, mass. Color, charge, and whatever else don’t count, because that’s misunderstanding gravity: gravity never could have acted upon those things, because they’re not the same kind of thing gravity acts upon (unless you mean something very different by gravity). It’s only in a trivial sense, very different than what is used in biology, that these properties are free-riders. Selection free-riders, however, do make sense, because the things acted upon and the free-riders are of exactly the same kind: namely, they are traits.
The proper response to what you’re calling a gravity free-rider (which would include all kinds of properties, like being a 2kg copy of Origin of Species as opposed to a 2kg copy of Hamlet: gravity acts on both identically!) would be to point out that gravity never could have acted on these properties, because they’re not mass. But you can’t make that move when you point out that it’s the color that’s the free-rider and not the insulating properties. It easily could have gone the other way around (and very well may in some other species). But if you say that the “gravity free-riders,” such as the massless gluon, or the color, or whatever, could have been acted upon by gravity, then you mean something very different by gravity.
It seems to me you’re drawing a modal distinction (one I see as special pleading). In the physics example, we have a class of things, including gluons, photons, and possibly neutrinos, that could have been massive in our world but the evidence indicates are not. In the biology example, we have a class of things, certain traits in certain species, that could have been adaptive but the evidence indicates are not. In the biology case the other possible worlds where these traits could have been adaptive are very close. In the physics case, the other possible worlds where gluons could have been massive are farther away (perhaps). I don’t see how this possible world distance has much relevance for Fodor’s argument. Again, seems like special pleading — to the particle physicist, these other possible worlds are of interest and pretty important.
Another point to make here is that part of Fodor’s argument is that NS is not “lawlike” enough for his taste. But from your own comments here it should be clear that NS is more generally applicable (has wider scope) than gravity, since as you say, adaptive traits can be anything, including qualities like color, taste, sound, etc. NS would also apply in worlds where gluons have mass, all else equal.
More tomorrow, maybe.
Talking about possible worlds in this context just invites confusion. It’s this very world in which what are sometimes free-riders are sometimes actually acted upon by selection, and vice versa. In one species, brownness of fur is a free-rider, in another, it’s selected for. That’s the whole point! That’s why free-riders are interesting in biology, and don’t make sense in physics. You seem to think I’m saying that, in a particular species, in a particular case, the brownness could have been different (a free-rider when it was actually selected for). What I’m saying is that, across species, it sometimes is actually different (that’s the modality of interest here, which is totally uncontroversial).
If this were otherwise, there would be no need for experimentation (or whatever else) to see, in a particular case, which was the free-rider, and which was not. If brownness was always selected for, and never a free-rider (which is what Fodor, rightly or wrongly, would want to call a law of selection), we would not need to check, no matter what the species.
In that case I think the distinction you are drawing between the physics and biology is completely arbitrary and false. If you assume from the outset that physics is privileged somehow, of course after working through the machinations you will conclude that the physics example is special. It’s called begging the question.
In one case we have a gluon, which happens to be a particle that has zero mass. It could have been a particle with mass (for instance, the particular particle in question could have been an electron). This particle doesn’t have mass, so it’s not acted on by gravity. In the other case we have a non-adaptive trait, that has a zero selection coefficient. It could have had a fitness differential, but it doesn’t, and so it’s not acted on by selection. The modality in this world is the same. Gravity doesn’t act on particles with mass of zero, and selection doesn’t act on traits with a selection coefficient of zero. Yes, there’s only one type of thing that gravity acts on, and that is mass. And there’s only one type of thing selection acts on, and that is a fitness differential. Hence both can be free-riders, since both can be coupled with things that are acted upon in the respective theories.
You aren’t responding to what I’ve said. I didn’t say that physics is privileged somehow; I just characterized physics. If you think I’ve mischaracterized it, then say why.
The point, once again, is that traits with, as you put it, fitness differential zero, have non-zero fitness differential when they occur in other organisms, in this world. Things with mass zero have mass zero everywhere in this world: they don’t have non-zero mass when they occur elsewhere in this world. I don’t know how else to put this.
There is no free-rider problem in physics, although there is the problem of determining whether something has mass or not. Once that’s determined, we know immediately whether gravity will act upon it, or not.
There is a corresponding problem of determining whether something’s a trait or not. Once that’s determined, there is the further problem (unlike physics) of determining whether that particular trait was selected for, or not. In another organism, the answer can be different. Not so for physics.
I still think you’re splitting hairs. Particles correspond to traits in this analogy, and particles can be massless or massive.
It’s easy to come up with a different analogy if you want. Consider a conglomerate rock, made of two different iron oxides. One of the iron oxides happens to be magnetized; the other does not. The rock exerts a magnetic field, and when it’s moved it can move other magnetic substances. Which iron oxide is responsible for the magnetic field? Both could be magnetized, and in other rocks one or the other or both often are. But in this rock you have to do experiments to figure out which one is magnetized (and in this case it is only one). The other, non-magnetized iron ore is a free-rider.
You could come up with obvious similar analogies with charged bars, etc.
I feel like I’ve beaten this point to death here, so I think this will be my last post.
It’s fine if particles correspond to traits: they come in two types. Gravity will act on those with mass, not on those without mass. Traits come in lots and lots of types: brown fur, insulating fur, etc. Each of those types can, in some instances, be acted upon by selection, and in some instances, be free-riders. It doesn’t do to say that traits only come in two types: the ones with non-zero fitness differential and the ones with zero fitness differential, and selection acts only on the former. We can label them as such later, at least for some traits, by experiment. But there is no amount of immediate observation that will tell you that brown fur, in some instance, is of type non-zero differential, the way you can immediately observe something’s mass, or magnetic field, and then let your theory tell you what to expect from there.
In the example you just gave, once again, we know from theory that the magnetized ore will exert a magnetic field, the other will not. Nobody will say that this is a free-rider problem that needs to be taken seriously in any way similar to the free-rider problem in biology. I can’t imagine a physicist, or a chemist, taking seriously the idea that there are free-riders on gravity, or magnetic fields, or combustion. But every biologist knows that sometimes traits are, and sometimes they are not, acted upon by selection.
Well, I am a biophysicist, so I fill both bills, and I take free-riders seriously in physics and in biology. So you don’t need to imagine it, it’s a reality.
You summarize the difference between the biological and magnetic scenarios as “sometimes traits are, and sometimes they are not, acted upon by selection.” But in my iron ore example, the same is true. Sometimes a specific type of iron ore is magnetized, sometimes not. Here we have something that is directly analogous in the very sense you wanted earlier. The same iron ore can be magnetized in one rock but not in another (same ore, different time, place, and history). And since the two different ores in the rock are physically coupled, orienting one in a magnetic field automatically orients the other — it’s a free-rider. EM theory doesn’t tell you which ore is magnetized. You have to do an experiment to find out. And once you have found out for this rock, that information doesn’t help you a bit with a different rock made of the same two ores, since they may be magnetized differently (e.g., the free-riding could be swapped).
Here’s the difference: “Sometimes a specific type of iron ore is magnetized, sometimes not.” And you’re right, EM theory doesn’t tell you which one was magnetized, but it does tell you what happens when a sample is magnetized. You’re equivocating on sameness here: in your example, the “same” iron ore is admittedly not the same, because in one case it has a property (a physical property, namely, being magnetized) that it elsewhere does not. This is not analogous to the case of traits: the very same trait, with all the same physical properties, could be a free-rider in one population, and selected for in another.
This is obvious in the case of domestic breeding, right? We could explicitly have a person breed for one of two correlated traits, and have another person breed for the other. In such an example, we would have no way to tell who selected for what (supposing that the two traits are perfectly correlated), even though there was a fact of the matter, which we could only learn by asking the breeders. The resulting animals would be indistinguishable; no amount of experimentation could tell you which was which. If we imagine the similar scenario, perhaps two populations that get separated, and one gets selected for one trait, the other gets selected for the other trait, we would have the same result (physically identical organisms), except we might be able to find out, via some feature of the environment, which trait was selected for in the respective populations. And that would be a real biological puzzle that would be worth figuring out.
Your examples are nothing like this: nobody would take seriously the idea that gravity or magnetic free-riders are theoretically motivated, whereas correlated traits are a theoretical issue in biology. Perhaps Fodor’s conclusion about the existence of this issue is incorrect, but you’ve not refuted that idea by showing that free-riders show up all over the place in science. Your examples seem to either not take seriously the real theoretical issue of correlated traits (biologists don’t call them free-riders, right?), or they suppose that these so-called free-riders in other areas of science are to be taken seriously in the same way, which they clearly are not (unless other scientists are quite mistaken about the non-existence of gravity free-riders and the like).
In fact I think it is you who is equivocating on sameness — I was just using the same type of “sameness” as you. The two iron ores are the “same” other than being magnetized; the two traits are the “same” other than one being the cause of the selection. The traits are not the same in the absolute sense — in one population one causes a fitness differential, in the other population it does not.
Again, you are privileging physical theories and concepts a priori, by arbitrarily reifying components of physical models, but not allowing me to do the same for biological models.
I don’t accept your domestic breeding example, simply because we can indeed tell who selected for what by watching them. If, on the other hand, you meant to say that the selection pressure had been removed before we could see anything — well then I could make the analogous move and say that a human agent had similarly removed the magnetism of the ore. In that case, we’d have just as much trouble determining which ore was responsible for the alignment of the rock in a magnetic field.
And you’re right, neither biologists nor physicists use the term “free-riders” for effects that are correlated but not causal. Scientists use the term “artifacts” and investigate them with positive and negative controls.
Okay, I think I’m done here. I’m not privileging anything, I’m just pointing out what I take to be quite uncontroversial features of the points we’re discussing.
I don’t know what you mean by traits being not the same in “the absolute sense.” Two physically identical 2kg masses (even molecule-for-molecule duplicates) aren’t the same in some “absolute sense” if one smashes into my laptop and the other doesn’t. But that’s not a difference that makes a difference for physics: all physical (and chemical) forces will act on each exactly identically (given that they are physical duplicates), even though they’re not the same in some “absolute” sense.
And as to the domestic breeding, this is precisely what Dmitri Belyaev did with silver foxes. Watching what the selectors did does not tell you whether they were selecting for tail wagging, or reduced fear response, precisely because they’re correlated! Turns out it it was reduced fear response, but tail wagging came along for the ride. It’s a pretty simple case, often used in intro evolutionary biology courses, to explain the problem of correlated traits, and why you can’t always tell just by looking.
And I really don’t know what you’re talking about by saying I’m reifying components of models. Are magnetic fields things? Is mass a thing? Without getting into unrelated discussions about scientific realism, these seem like things to me; I mean, we have instruments to measure them. I suppose if you’re a nonrealist about mass and magnetic fields, your critique is valid, but then you’d certainly better be a nonrealist about traits and causal effects as well.
Thanks for taking the time to have this discussion; I’ve put up a post inspired by it over at the blog I contribute to:
We can continue there if you’d like; we’ve probably taken up enough space here!
You may take up as much space here as you like on this erudite, interesting and above all civil debate. But I shall subscribe to your feed anyway.
Since John doesn’t seem to mind ….
I’m conversant with Belyaev’s fox breeding experiments. And they do not help make your point. Selecting for fear response is not the same as selecting for tail wagging, and it should be fairly easy to tell the difference. For one thing, if the two traits aren’t correlated, then selecting for fear response will not, on average, change the frequency of tail wagging (and vice versa). Hence there necessarily must be a difference between the two activities. If nothing else, I could do my own experiment with Belyaev, where, say, I secretly slip in foxes with clipped tails and see how/if they get selected. There is a close analogy here with the magnetism example — you could watch me align the rock in a magnetic field, and unless I tell you or unless you do an experiment yourself, you cannot know which of the iron ores in the rock is magnetic and which one is free-riding. The rock will get aligned identically regardless of which ore is magnetic.
You seem so very surprised that physicists (or chemists, or biochemists, or what-have-you) would be seriously concerned about “free-riding”. But it is extremely serious, and it is probably the main obstacle to making valid scientific inferences. As I said, scientists routinely think very hard, go to great lengths, and spend large amounts of time to distinguish true causality from mere correlation — this is the very purpose for positive and negative controls.
In regard to “sameness”, taking a comment you made earlier:
“the `same’ iron ore is admittedly not the same, because in one case it has a property (a physical property, namely, being magnetized) that it elsewhere does not. This is not analogous to the case of traits: the very same trait, with all the same physical properties, could be a free-rider in one population, and selected for in another.”
In my view, these are indeed analogous. You see magnetization as a property of the ore (by EM theory); but having a non-zero selection coefficient (i.e., being causally responsible for a fitness differential) is likewise a property of a trait. The selective properties of a trait are a function of its environment (by theory). The same trait in a different species but in the same environment will have the same selection coefficient.
Clearly there are differences between these analogies, and between the physical and biological theories, but they are not the differences you are trying to draw. And more importantly, the differences you have pointed out (erroneously in my view) in any case do nothing to support Fodor’s conclusion, based on the existence of “free-riders”, that “natural selection doesn’t work”.
The point about bringing up the Belyaev example was precisely that the traits are correlated: you can’t change the example to say that they aren’t! And I contend that, given the fact that less-fear and tail-wagging were, in fact, correlated, you could not tell, just by watching, whether the breeders were selecting for tail-wagging or for less-fear. You’re right, you could run the experiment to find out by removing the tails, or injecting a chemical into the foxes to make them have less fear but with no tail wagging. But if you had two cases with populations of unaltered foxes, one in which breeders were selecting for tail-wagging, and one in which they were selecting for less-fear, you would end up with the same animals in each population, and you wouldn’t be able to tell which breeders had, in fact, selected for which trait, even though there was a fact of the matter (which you could find out only by asking the breeders). That’s why they bring up this case in classes, right? To illustrate that there really are free-riders (and in this case, it was tail-wagging).
You can change the example however you want, but selecting for less-fear and selecting for tail-wagging are not processes that you can distinguish by looking at them (breeders who pick out the best tail-waggers are, de facto, also picking out the ones with least fear, and vice versa, even though each may only be paying attention to one or the other), nor are they processes that will give you two different populations. This is a simple point: to use a Fodor-inspired example, if you pick out roses for their redness, but redness is perfectly correlated with length of stem, I can’t tell, just by watching, whether you’re picking the roses you are because of their redness or their stem length. Say it’s really redness in one case. Later, you decide you want to get roses with the longest stems. I watch again, and I still can’t tell whether you’re selecting for redness or stem-length: the process, and the results, are identical.
The free-riders in other sciences that you mention just seem to be instances of separating correlations from causes, which is a one-time project. Sure, a scientist might wonder whether heat (or whatever) causes some particular reaction, or is correlated to that reaction. Sure, it might be really hard to find out. But once you’ve found out, induction is possible; it’s not as if later on, or in another laboratory, what was once a cause will turn into a correlation, and vice versa. If that happened, there would be a genuine free-rider problem for chemistry, and we couldn’t make inductions. But there is a free-rider problem for biology, which is precisely why you can’t induce from particular cases of brown fur having been acted upon by selection to all cases of brown fur are (and will be) acted upon by selection: it’s always possible that the acted-upon trait will be a free-rider in a different context.
You’re right that having non-zero selection coefficient is a property of a trait. But properties are cheap. Another property of a trait is its being expressible in English in fewer than 13 letters. Similarly, a sample of water has the following properties: having a certain mass, having a certain number of molecules of a certain structure, having been blessed or not (i.e. whether it’s holy water or not), having been taken from a stream in Nebraska, etc. Obviously some properties are going to be “acted upon” by forces, or mechanisms, or whatever. Gravity works on the mass properties, maybe chemical forces (so to speak) work on the chemical structure, maybe priests work on the holy-water-or-not property (at least insofar as they believe they are dealing with holy water rather than regular water).
The suspicion that I am (and I think Fodor is) trying to articulate is that non-zero selection coefficient is kind of like having come from Nebraska. Given two samples of identical H20 (suppose they’ve been filtered), there is a fact of the matter about which one came from Nebraska, and which didn’t. You won’t be able to tell by the physical properties (just by looking) alone. And no physical process will act differentially on the two. But if you do some research, ask the right people, maybe you can figure out which is which. Similarly, in my example of duplicating Belyaev’s foxes so that one set of breeders selects for tail-wagging, and the other selects for less-fear, you won’t be able to tell by examining the foxes (or the breeders) which trait had non-zero selection coefficient. You could if you asked the breeders, or course, because they know what they respectively took as the cause. Then you could know the selection coefficient. But selection coefficient is not a property to be found in the actual organism (we have two physically-identical populations, but one has zero selection coefficient for tail-wagging, and one has non-zero selection coefficient for tail-wagging), so it’s hard to see how a physical process could act on this difference.
I did say earlier that perhaps Fodor draws the wrong conclusion from this. Still, I think it’s interesting, and does show an interesting difference between biology and at least the physical sciences. And like I said before, you haven’t refuted Fodor by showing that there are free-riders (or even anything like free-riders) everywhere in science, which I think was your original intent.
I didn’t change the Belyaev example — I simply pointed out that causes often have more than one effect. Just because (1) selecting for tail-wagging and (2) selecting for fear response can both produce the same effect E, does not mean that (1) and (2) only produce effect E. (1) can also produce effect A, and (2) can also produce effect B. So I can distinguish between the two causes, not from effect E, but from the other effects A and B. For instance, if I were to watch Belyaev, and note that he never looks at the foxes’s tails. Hence, your claim that:
“Watching what the selectors did does not tell you whether they were selecting for tail wagging, or reduced fear response, precisely because they’re correlated!”
is in general false.
The problem of “free-riders”, tail-wagging vs fear response, red roses vs stem length, etc., is nothing more than an evolutionary case of a ubiquitous problem in science, of having an effect that could be due to multiple causes (causes which may or may not be mutually exclusive). In physics, magnetism of one ore or the other being responsible for alignment of the rock, for example.
You further claim that “separating correlations from cause” in other sciences is a “one-time project”. This is also false. Take the magnetism example. I can discover, by experimentation, which ore in my rock is responsible for the alignment in a magnetic field, and which ore is just getting aligned by being physically correlated with the magnetic ore. Then later, in another lab, a rock with the same composition may have the other ore magnetized. Here we have a case that “later on, or in another laboratory, what was once a cause will turn into a correlation.” Now you may object that a property of the rock has changed, and I’ll agree. But similarly, a trait that is selective in one environment can be non-selective in another. But only if the environment has changed. Remember, selection is a function of the environment — that’s inherent to the theory (dependence on external variables is generally the rule throughout physics also). If, rather, you keep the experimental setup the same, the same trait will be selective in the same (or even similar) environment. So you can use induction just as well in the biological case as in the physical case. Of course you will see differences between physics and biology if you keep the variables constant in physics but allow them to change in biology — this is why I’ve charged you with a priori privileging physics.
Gravity never could “pull” on things without mass, whereas free riders in biology could be operated upon.
No — free-riders in biology are not operated on by natural selection, but they can be confused by us as being causally responsible for a fitness differential. The analogy is almost perfect in this regard. F&P are conflating what is causally responsible for selection with the effects of selection.
Consider neutrinos, where it is certainly not confused to ask — and is still an outstanding question in physics — whether gravity is pulling on them in fact or is only “pulling” on them via “local confounds”.
You’re missing the modal point here. There’s a difference between saying that free-riders are not operated on, and saying that they could not be operated on. A trait that’s a free rider in a particular case could have been causally responsible (and thus not a free rider) in a different case, under the very same theory. That’s why you have to do experiments (or whatever else) to figure out which trait was responsible, and which was the free rider.
I don’t know much about physics at all, but if neutrinos have non-zero mass, then they are candidates for being affected by gravity. If they are massless, as is the orangeness of the basketball, they aren’t even candidates (at least given what I know of Newtonian mechanics). If someone is confused, wondering whether it’s the massless property of orangeness or the mass that is being operated upon by gravity, given that the two always (let’s say) go together, we point out that being orange could never have been operated on by gravity, according to Newtonian mechanics. You can’t say the same thing when you wonder about which trait is a free rider: in one context a trait is a free rider, in another it is being operated upon.
It is an open question whether neutrinos have mass or not. There is some evidence that they do, but that is beside the point. The point is that it is a valid question to ask whether neutrinos have mass, and, if so, what the value is. The analogous problem in selection theory is whether there is a trait that results in a fitness differential, and if so, what the value is.
The property of “being orange” is a bit too obvious. What if we observe a new particle from the large Hadron collider, and we see that its trajectory forms a parabola? Obviously the particle was acted on by a force, but was it because the particle has a large mass, or because it is charged? Or both, or something else? Clearly, if our theories are correct, then the charge cannot be operated on by gravity, and neither can the mass be operated on by an electromagnetic force. But the particle could be operated on by either, depending on its characteristics, which are a priori unknown to us. If the particle is charged but with extremely small mass, then the mass of the particle is a “free-rider”.
Do you see how close this analogy is? Here we have a charged mass that is accelerated in an EM field — lets say in the absence of a gravitational field. The acceleration of the mass is solely due to the charge that the mass is associated with.
In biology we have trait B that is swept to fixation because it is genetically linked with trait A — lets say trait B has no causal effect on fitness. The fixation of trait B is solely due to the selection coefficient of trait A that trait B is associated with.
It is absurd to think that either of these scenarios is a “counterexample” to the underlying theory. At least from the perspective of a scientist — the philosophers obviously see things a bit differently.
I still don’t see the cases as analogous.
In your example, trait B has no causal effect on fitness, but trait A does, and the two are coextensive. The “force” in question, in fact, acted on A, but could have acted on B; we determine, through whatever means, that it was A that did the work. Both are candidates for being acted upon by the one force, which is why we have to do some work to figure out which is, in fact, the free-rider (it’s not a priori).
In your particle example, the mass never could have been acted on by the force in question, namely, EM. And, if all we had was Newtonian mechanics (NM), your situation would in fact be a counterexample to the claim that NM explains all motions. In that case (which is what actually happened in the history of science), we must look for an additional force, something more than gravity. Then, when faced with moving particles, we can know, a priori (assuming we are right about gravity and EM forces), what possible contributions of mass and charge could result in the observed movements; we just have to do some experiments to find out which configuration we are faced with. And if nothing fits, then we might go looking for yet another force. The mass is not a free-rider in any sense here, because we have two forces acting on two different kinds of property.
This is totally different than the trait scenarios: either trait is a candidate for being acted upon by the very same force, and NS alone won’t tell you which one. Which is not to say (which Fodor emphasizes) that there is no fact of the matter about which one it is, but rather that you have to go beyond NS to find out (as in doing experiments, etc.).
You’re losing the point splitting hairs. I don’t see why you think the two scenarios are so different — they are analogous in exactly the ways that matter. Why does it matter that in the physics case there are two forces?
For the analogy to be inapplicable, the differences you mention should support Fodor’s argument that NS “doesn’t work”. Yet in both the biological and physics case, you have to do control experiments to figure out what is causally responsible for the observed phenomenon. In the physics case you also have to “go beyond” the theory, since neither classical gravity nor EM theory can tell you why the particle accelerated.
I also wonder why you mentioned that “situation would in fact be a counterexample to the claim that NM explains all motions”. Are you implying that natural selection is supposed to explain all evolution? Are you implying that Fodor thinks this? If so, that would go a long way to explaining his confusion.
FWIW, the Price Equation can be considered the “law” of selection. It is arguably more generally true than any physical “law”.
Thanks for the link. I think Sober’s right to point out that there may be a lot more to theoretical evolutionary biology than he (Fodor) thinks, and that the mathematical (and other) models used there are more general than he (Fodor) thinks.
Comments are closed.