Wednesday, April 29, 2009

More on Teleology in Evolutionary Biology

I’ve been corresponding via email with a fellow evolutionary biologist (who shall remain nameless). I thought that some of her/his comments might be useful or interesting to those who read this blog, as they have a direct bearing on the "problem" of purpose (i.e. teleology) in evolutionary biology.

My correspondent's comments and questions are in block quotes:
I’ve been following the ‘Survival of the Sickest’ thread at Uncommon Descent, and have some comments on it and on your essay on the (un)reality of adaptations.

First, a major area of agreement between us is that the original post is fatally flawed by:

1) the assumption that “Darwinism” implies “constant progress”, and

2) failure to understand that fitness is always defined relative to a particular environment, and that environments change over time.

Now, on to some points of disagreement:

You mentioned Gould and Vrba’s 1982 paper on exaptation, and wrote:

"The reason is quite simple: if (as Gould, Lewontin, and Vrba argue) adaptation isn’t legitimately part of what evolutionary theory is about, then the whole idea of “design” and “function” is read completely out of evolution, leaving only descent with modification."

I have been very strongly influenced on this topic by Warren Allman, director of the Paleontological Research Institute here in Ithaca. He asserted that all adaptations should be considered to be exaptations. His rationale for this assertion was that the term “adaptation” has built into it an assumption of teleology. Literally translated, the word "adaptation" means “toward usefulness”. It’s the “toward” part that is the problem. As you and I both understand it, evolution (including natural selection, but not artificial selection) does not tend toward anything. It has no goal as far as we can tell. Ergo, it builds on what has gone before, but without any specific goal “in mind”.

This is why “exaptation” expresses better how we understand natural selection. It builds “away from” non-functionality (or even away from previous functionality), but never really “toward” anything at all as far as we can tell. And, if Sewall Wright’s “shifting balance” theory is a reasonable model of evolution, then it never really “arrives” anywhere at all, since the “goal” is constantly shifting anyway.
I’m wondering why you think that Gould and Vrba regard adaptation as being outside the legitimate scope of evolutionary theory. My take on the paper is not that they regard the concept of adaptation as illegitimate, but just that it has been typically construed too broadly and should be broken down into the categories of true ‘adaptation’ and ‘exaptation’, where they define a true adaptation thusly:

“Following Williams, we may designate as an adaptation any feature that promotes fitness and was built by selection for its current role.”

The problem I have with this definition is the inclusion of the words “promotes” and “for”. “Promotion” means exactly what it says: “motion towards” something. Ergo, using this word immediately suggests teleology, and as I have pointed out above, teleology cannot be a valid assumption in the origin of the products of evolution at any level. This is not because including teleology allows for “a divine foot in the door” (c.f. Lewontin, 1997), but rather because it requires that the “plan” for the teleological process must exist prior to the coming into being of that process. When we do things, this assumption is perfectly valid, but when something happens in nature, such an assumption is entirely unwarranted. Where, in nature, could such a pre-existing plan exist?

As for the word “for, I always point out to my students that teleological explanations virtually always reduce to sentences that include the phrase “in order to”. This can be shortened even further to “to” (leaving out the “in order”). However, the entire phrase “in order to” can be replaced with the word “for” without changing its meaning. Ergo, the definition quoted above is still irreducibly teleological, and therefore includes an assumption that we should not make in evolutionary biology.

In my paper on the evolution of the capacity for religious experience, I began with a succinct definition of “adaptation”, from which I lay out four criteria that a characteristic (i.e. a “trait”) must meet to be considered a genuine adaptation.

An evolutionary adaptation is any heritable phenotypic character whose frequency of appearance in a population is the result of increased reproductive success relative to alternative versions of that heritable phenotypic character.

Here are the four criteria that I believe must be met for a characteristic to be considered to be an adaptation:

1) An evolutionary adaptation will be expressed by most of the members of a given population, in a pattern that approximates a normal distribution;

2) An evolutionary adaptation can be correlated with underlying anatomical and physiological structures, which constitute the efficient (or proximate) cause of the evolution of the adaptation;

3) An evolutionary adaptation can be correlated with a pre-existing evolutionary environment of adaptation (EEA), the circumstances of which can then be correlated with differential survival and reproduction; and

4) An evolutionary adaptation can be correlated with the presence and expression of an underlying gene or gene complex, which directly or indirectly causes and influences the expression of the phenotypic trait that constitutes the adaptation.

I would now modify criterion #4 to state that such genes/gene complexes must be shown to have been conserved, relative to other sequences in the genome. However, one must keep in mind that such conservation, while necessary, is not sufficient. As we know now, some sequences are conserved, but can be knocked out, with no discernible effect on phenotype. Ergo, to fully satisfy criterion #4, a characteristic must be shown to be associated with a particular gene or gene complex, the knocking out of which can be shown to have significant negative effects on fitness.

Obviously, this means that a great many characteristics that we observe in living organisms will not qualify as adaptations. I believe that this is fully justified, following Williams’ assertion that the concept of adaptation is “onerous” and should only be resorted to “in the last resort”. It is only by doing so that we may avoid the otherwise almost inevitable pitfall of appealing to teleology in our explanations.
Gould and Vrba close their paper with this:

“The argument is not anti-selectionist, and we view this paper as a contribution to Darwinism, not as a skirmish in a nihilistic vendetta. The main theme is, after all, cooptability for fitness. Exaptations are vital components of any organism’s success.”

There’s that nasty little word “for” again! Fitness is immediately measurable as relative differential reproductive success, but “adaptation” can only be legitimately inferred retrospectively. We can’t say that something is a genuine adaptation until it already is, and this seems to me the kind of logical circularity that has also plagued Herbert Spencer’s phrase “survival of the fittest”. If we stick to the four criteria listed above, we will rarely fall into the trap that teleological thinking always sets for us.

Also, you later wrote the following, which seems to acknowledge that Gould and Vrba did regard adapation as a legitimate part of evolutionary theory:

“Yes, indeed, except that I believe that Gould, Lewontin (and later, Vrba) were, like Darwin, unwilling to take their principles to their logical conclusion: that adaptations (like species) are a figment of the human imagination, and do not actually exist in nature (or, to be even more precise, do not have to exist in nature).”

What I meant by this is that the only way we can actually “detect” the presence of adaptation is by inferring it. In that sense, adaptations are not “primary” characteristics; that is, characteristics that can be directly observed (such as differential reproductive success). Rather, such “secondary” characteristics must be indirectly inferred. In that sense, they are indeed “imaginary”; we must “imagine” that they exist (as the result of our application of inferential logic), as we cannot observe them directly.
Am I missing something? Are you trying to say that although Gould and Vrba regarded adaptations as real, they nevertheless thought they should be excluded from evolutionary theory?

No, I’m saying what Williams was saying, only I’m saying it more strongly and consistently: that we should never include any hint of teleology in our explanations, as such inclusion includes the biological equivalent of that old bugaboo of physics: “action at a distance” in physics is the equivalent of “goals preceding causes” in biology.

When I reread Williams’ famous 1966 book, Adaptation and Natural Selection, which supposedly reads teleology out of evolutionary biology, I was astonished to find it shot through with the same kind of teleological reasoning that he was supposedly trying to eliminate. I think I could find all the “hidden teleology” in Williams because I have spent so much time debating with ID supporters. They are the ultimate teleologists, and can always find where we have subtly woven teleological assumptions into our biology.
Finally, you wrote:

“To be as clear as I can, I believe that asserting a position of “metaphysical materialism” is just that: a metaphysical, not a scientific assertion. Confusing metaphysics with science is nearly as pernicious as confusing “ought” and “is”. The former makes for questionable science and the latter makes for questionable ethics.”

I would agree that science has no say on metaphysical questions that don’t have observable consequences (although I would argue that even then, Ockham’s razor should cause us to prefer simpler metaphysical systems to needlessly complex ones). However, some metaphysical assertions do have observable consequences. For example, I consider the existence of the Young Earth Creationists' God to be a metaphysical assertion that has nevertheless been decisively falsified by science.

I agree, but the same cannot be said for the more subtle versions of teleology found in Michael Behe or William Dembski's works. Their books (especially Dembski’s) present a much more subtle and less easily refuted version of teleological explanation, one that is easily reinforced by our own unwitting resort to teleological explanations.

Evolutionary adaptation is where the rubber of both evolutionary theory and ID hit the road.
Now on to your essay “Are Adaptations ‘Real’?”

You wrote:

“...although there are characteristics of organisms that are correlated with relatively high reproductive success (and would therefore be considered by most evolutionary biologists to qualify as “adaptations”), it becomes problematic to decide exactly which of those characteristics are the “real” adaptations and which are merely ‘accidental’”.

The problem, of course, is the words “real” and “accidental”. If we are genuinely dedicated to rooting out teleology in all of our explanations of the origins of biological objects and processes, then all adaptations are “accidental”, in the sense that they are all unplanned. We perceive them as having “functions” because our naive viewpoint of reality is always teleological. We can think non-teleologically only with very great difficulty. It’s like special relativity or quantum mechanics. We have to twist our minds to be able to even begin to conceive of them, and even then we constantly slide back into our naive (and unwarranted) views of reality.

True, if by “accidental” adaptations you mean exaptations. But while it may sometimes be difficult to tell whether an adaptation is “real” or “accidental”, that is not evidence that “real” adaptations don’t exist. Indeed, the only scenario I can envision in which “real” adaptations would not exist would be one in which every fitness-enhancing feature was an exaptation.

But that would mean, among other things, that every incremental improvement to the eye would have to have been the accidental result of changes that were selected for some reason other than improved vision. That seems far-fetched to me. Am I misunderstanding your position?

It’s not that that every incremental improvement to the eye would have to have been the accidental result of something, it’s that every incremental change to the eye would have had to originate accidentally, but then increase in frequency as the result of differential survival and reproduction. If we think the way you worded it (and we almost always think that way), then the teleological trap is that all of the incremental changes are somehow “predestined” and that complex eyes must be the inevitable result.

But this just plays into the hands of intelligent Design supporters. When we argue that “half an eye is still adaptive” we unwittingly include the assumption that “half an eye” is just that: half of what will ultimately evolve by natural selection. But our knowledge of the natural history of vision has shown us over and over again that “half an eye” is the whole thing in many cases. We can only say that the eyes of, say, flatworms, are “half an eye” because we already know that such a thing as a “whole eye” exists in cephalopods and vertebrates. We have to disabuse ourselves of the idea that any characteristic is only partially the whole deal. All characteristics of all organisms are the whole deal for those organisms, period, end of story, that’s all She wrote. Anything else contains the beginnings of teleology, and that way lies error, endlessly compounded.
We now have the ability to selectively delete individual characteristics from many different organisms. This makes possible something that natural selection does not: the precise determination of the selective “value” of particular characteristics. This has already been done, and the surprising outcome has been that even some gene sequences that were thought to have been very important in selection (due to having been “conserved” over deep evolutionary time) are apparently insignificant or useless. We know this because knocking them out of the genome has no discernible effect on the survival or reproduction of the “knock-out” progeny.

Precisely my point, above.
That interpretation seems to depend on the hidden assumption that the environment hasn’t changed significantly in the recent history of the organism, and that the experimental environment is fully representative of the historical environment over the entire time during which the features in question evolved. In the case of knocked-out sequences that have no apparent effect on fitness, how sure are we that the experimental environment is fully representative in this way?

No, but to assume that we are making the opposite mistake - assuming that some characteristic really has some function, even if that function is entirely unobservable - is once again to fall into the “teleology trap”. This is essentially the same argument that ID people make about “junk DNA”. Just because we haven’t found any function for it, doesn’t mean that all of it has no function. They argue that all of it must have some function. They are, like the evolutionary biologists for whom Williams, Gould and Lewontin, and Gould and Vrba wrote their warnings about, assuming teleology in evolution: they are, in a word, “pan-adaptationists”.
As a hypothetical example, imagine a bacterial DNA sequence that is expressed only during the formation of spores to protect the organism during periods of extreme environmental conditions. Knock out the sequence and test the viability of the resulting variant. If the experimental environment doesn’t include the extreme conditions that induce spore formation, the organism will never attempt to express the knocked-out sequence, and so its absence will not be noticed. If the experimenter concludes that the sequence is insignificant or useless, she is mistaken.

True, but I would strongly prefer that adaptation be considered a “diagnosis by exclusion” rather than our first and most important resort. By focusing on adaptation and natural selection, we teeter on the edge of the “teleology trap” and often (maybe even usually) fall in, despite our best efforts to avoid doing so.


As always, comments, criticisms, and suggestions are warmly welcomed!


Labels: , , , , , ,


At 4/29/2009 08:03:00 PM, Anonymous Dave Wisker said...

Hi Allen,

Just a note to say how much I like this essay, and to add one comment:

This has already been done, and the surprising outcome has been that even some gene sequences that were thought to have been very important in selection (due to having been “conserved” over deep evolutionary time) are apparently insignificant or useless. We know this because knocking them out of the genome has no discernible effect on the survival or reproduction of the “knock-out” progeny. I think we have to be careful with the conclusions drawn here. These knockout experiments are done on laboratory mice, to the best of my knowledge. As such, they have a 'genetic background' suitable for populations of mice bred for laboratory environments. That is, they possess coadapted gene complexes assembled over many, many generations of strong selective breeding for that environment. So, it seems to me if you knock out a highly conserved sequence, which most likely arose with a completely different set of coadapted gene complexes in the wild, the fact its removal has no apparent effect may be deceptive. I'd like to see some attempts at knocking out the same sequence in mice from strains with the coadapted gene complexes from the wild, and see if the fitness effect is different. Of course, there are practical problems with this, but I don't think we can conclude the sequence has no value on fitness until we remove it from the genetic backgound from which it originally arose.

On a different note, I want to thank you for your support over on UD. I suspect, since I have been posting on ARN (under the name "KC") and occasionally guest blogging on the Panda's Thumb under my real name, that they see me as having a particularly virulent form of ID-Critic cooties. ;)

At 4/30/2009 09:39:00 AM, Blogger Allen MacNeill said...

Here's a comment from my correspondent (the one whose questions form the structure for this post) that came in via email:

I think you are actually intertwining three separate arguments, which is perhaps why it was hard for me to understand your position at first. One argument is that we should excise all teleological language from discussions of evolution, including words like ‘adaptation’, ‘promote’, and even ‘for’, since all of these can be construed as suggesting teleology. Your second distinct argument is not for evicting the word ‘adaptation’ from evolutionary biology, but rather only for tightening the criteria by which adaptations are recognized. Third, you argue that adaptations can only be identified retrospectively and that this renders the concept of adaptation suspect.

Regarding the first argument, I think that ridding evolutionary biology of what you consider to be teleological language is a forlorn hope, because words like ‘promote’ are ubiquitous in all of the sciences, not just evolutionary biology (e.g. “excess use of fertilizer promotes algal blooms” or “slow cooling promotes the growth of large crystals”), even when the meaning is clearly neither intended nor understood as teleological. Indeed, non-teleological deployments of teleological words are found not just in all of the sciences, but everywhere that language is used.

Teleological language comes naturally to us, as evolved social creatures, and we find ourselves using it metaphorically even when the ideas we are expressing are not actually teleological. We do the same thing with spatial language (“he’s in over his head”) and even food language (“she bit off more than she could chew”). Note the embedded spatial metaphors in the words “understand”, “oversee”, “circumvent”, etc. We learn very early when and when not to take these literally, and I don’t see any reason that we can’t do the same with teleological language.  

Avoiding teleological language also results in some incredibly awkward locutions. Try recasting the following sentence in non-teleological language to see what I mean: “In arctic hares, evolution has selected white fur for purposes of camouflage.” For my money, it is easier to guard against teleological thinking while continuing to use teleological language than it is to recast everything in unnatural and awkward teleology-free terms.

Also, the point of Gould and Vrba’s paper was to provide a finer-grained “taxonomy of fitness” by distinguishing adaptations from exaptations. To label all changes favored by selection as ‘exaptations’ is to erase Gould and Vrba’s useful distinction. (And returning to my previous point, note how natural it is to use the word ‘favored’ here without any implication that selection has conscious preferences).

Last, to say that “adaptations aren’t real” is misleading. We agree (I think) that the word as actually used by biologists refers to something real and nonteleological. It’s just that you and Warren Allman think that a different term, ‘exaptation’, is a more appropriate label for this concept.

I am more sympathetic to your second argument for tightening the criteria by which we identify true adaptations. However, I would stress that a trait that fails to meet your proposed criteria may nevertheless be a true adaptation – just one that we haven’t conclusively identified as such.

You state your third argument this way:

Fitness is immediately measurable as relative differential reproductive success, but "adaptation" can only be legitimately inferred retrospectively. We can't say that something is a genuine adaptation until it already is, and this seems to me the kind of logical circularity that has also plagued Herbert Spencer's phrase "survival of the fittest".

Is this really a problem? The concept of a ‘most recent common ancestor’ also can’t be applied except in retrospect, but does that make it circular or any less real?

A final source of confusion for me is your formal definition of an adaptation:

An evolutionary adaptation is any heritable phenotypic character whose frequency of appearance in a population is the result of increased reproductive success relative to alternative versions of that heritable phenotypic character.

This definition is broad enough to encompass exaptations. You’ve erased Gould and Vrba’s distinction again, but this time by lumping everything together back under the term ‘adaptation’ instead of ‘exaptation’. Isn’t that what you are trying to avoid?


P.S. Nice takedown of StephenB on methodological naturalism. I was going to suggest mentioning Buridan, but I see that you’ve already done that and more. Stephen’s argument makes about as much sense as insisting that Newton wasn’t a scientist because the term ‘scientist’ wasn’t used before 1833.

At 4/30/2009 12:42:00 PM, Anonymous mauka said...

Hi Allen,

I thought I should correct something in the OP. You credit me as a "fellow evolutionary biologist", but I'm actually an engineer (and a happy exception to the "Salem Hypothesis") with a strong interest in evolution.

At 5/07/2009 04:05:00 PM, Blogger ghostbeef said...

I have some questions:

1. Provide one example of a change in the gene expression level of a developmental control gene results in a new trait or loss of a trait.

2. Describe the anatomical and physiological changes to love-finned fish that accompanied the evolutation of partially land-dwelling tetrapods. (Include Tiktaalik, Icthyostega, and Acanthostego)

3. Describe the rold of human-created waste in the creation of low-oxygen "dead zones" in oceans, particularly near rivers. Assuming that hthis is a permenent condition, what kinds of organisms are more likely to expand via adaptive radiation in this environment?

4. Provide on example of a behavior or other mechanism that prevents cheating in an evolved social cooperation system.

5. Describe the hisotry of measles in the human population. focus on where it came from, and whether its virulence became attenuated.

6. Describe a common explanation for cause of the mass extinction even at the end of the Cretacious period. Which animal groups where prevalent prior to the extinction even, and which groups expanded after the event.

Just an enthusiast


Post a Comment

<< Home