Intelligent design proponents make a negative argument for design. According to them, the complexity and diversity of life cannot be accounted for by unguided evolution (henceforth referred to simply as ‘evolution’) or any other mindless natural process. If it can’t be accounted for by evolution, they say, then we must invoke design. (Design, after all, can explain anything. That makes it easy to invoke, but hard to invoke persuasively.)
Because the ID argument is a negative one, it succeeds only if ID proponents can demonstrate that certain instances of biological complexity are beyond the reach of natural processes, including evolution. The problem, as even IDers concede, is that the evidence for evolution is too strong to dismiss out of hand. Their strategy has therefore been to concede that evolution can effect small changes (‘microevolution’), but to deny that those small changes can accumulate to produce complex adaptations (‘macroevolution’).
What mysterious barrier do IDers think prevents microevolutionary change from accumulating until it becomes macroevolution? It’s the deep blue sea, metaphorically speaking. IDers contend that life occupies ‘islands of function’ separated by seas too broad to be bridged by evolution.
In this post (part 2a) I’ll explain the ‘islands of function’ metaphor and invite commenters to point out its strengths and weaknesses. In part 2b I’ll explain why the ID interpretation of the metaphor is wrong, and why evolution is not stuck on ‘islands of function’.
Read on for an explanation of the metaphor.
The ‘islands of function’ metaphor
The ‘islands of function’ metaphor is a variation of another metaphor, the ‘fitness landscape’. If you’re unfamiliar with the concept of fitness landscapes, I encourage you to do some Googling before reading on.
For those who are familiar with fitness landscapes, a brief review. Imagine a three-dimensional landscape, similar to a terrestrial landscape. There are mountains and depressions, ridges and valleys, plains and plateaus. An organism occupies a particular spot on the landscape. Nearby spots represent organisms that are similar, but with slight changes. As you move further away from the spot, in any direction, the organisms represented become less and less like the original organism.
Evolution can be visualized as a journey across such a landscape. Individual organisms don’t move, but their offspring may occupy different nearby spots on the landscape. So too for their offspring’s offspring, and so on. Thus successive generations trace out a path (or paths) on the fitness landscape as changes accumulate.
Clearly, not all paths are possible. Many mutations are deleterious, causing their possessors to die young or to otherwise fail to reproduce. Paths going through such points on the landscape will end abruptly. Other mutations are beneficial, neutral, or only slightly deleterious. Paths going through those points may continue.
Now let’s bring in the third dimension, height. The height of a point on the landscape is an indication of the fitness of the corresponding organism, where fitness equates to the organism’s ability to survive and reproduce. Greater heights correspond to higher fitness, lower heights to reduced fitness. Offspring that move downhill from their parent(s) are less fit, and therefore tend to leave fewer offspring of their own. Offspring that move uphill from their parent(s) are more fit and tend to leave more offspring. Over time, then, a population tends to shift in an uphill direction as the offspring become fitter.
Eventually the population may reach the tip of a peak and get stuck there. From the peak, movement in any direction results in less fitness. Thus the mutants will tend to die off and the population will remain at the tip of the peak.
So far we’ve been imagining a dry landscape. Now suppose that it rains for 40 days and 40 nights. The rain fills up our landscape, forming a vast sea. Only the mountain tops remain above the water as islands – the ‘islands of function’ that IDers are so fond of.
Our populations occupy the islands. Sea level indicates the minimum fitness at which mutants remain viable. Small changes will create viable descendants at different spots on the island, though the population as a whole will gravitate toward the high spots. Larger changes will put the mutants underwater, where they will die out.
The idea, according to ID proponents, is that populations remain stranded on these islands of function. Some amount of microevolutionary change is possible, but only if it leaves you high and dry on the same island. Macroevolution is not possible, because that would require leaping from island to island, and evolution is incapable of such grand leaps. You’ll end up in the water.
There is some truth to the ‘islands of function’ metaphor, but it also has some glaring shortcomings that ID proponents almost always overlook. I will mention some of the strengths and shortcomings in the comments, and I know that my fellow commenters will point out others.
I may add them to the OP as they come up in the comments. If I do this, I will note that I am doing so and I’ll include a link to the place in the comments where each one is discussed.
Have at it!
Mung, please try the following exercise.
See? That was easy.
If there are any words you are still having trouble with, I would recommend an eight-grade geometry textbook.
BTW you do realize that you are providing a live demonstration of Mike’s point re the lack of understanding displayed by IDers.
ETA: sorry keiths, you beat me to it…
1) Show evidence of guidance.
2) If evolution needs to be “guided” then the implication is that there is a target that it needs to be guided to.
3) Show evidence that the lifeforms we see are “required” to be in the state they are.
Mung:
Where? I haven’t seen any evidence that you even understand the ID side of the ‘islands of function’ argument, much less my counterargument.
No, Mung, an ad hominem involves attacking your opponent in an irrelevant way in order to discredit his or her argument. I couldn’t do that even if I wanted to, because you haven’t succeeded in making an argument.
The reason I brought up your dishonesty is to not to discredit your nonexistent argument, but to demonstrate this: you are so eager to disagree for the sake of disagreement that you will actually lie about your opponent’s position in order to have something — anything — to disagree with.
That’s pathetic, Mung. Why don’t you go off, learn what the issues are, and then come back when you can make an actual, relevant argument that doesn’t require lying?
Meanwhile, I see that you are still completely confused about ‘islands of function’. Regarding the McLaughlin paper, you wrote:
Mung, the ‘islands of function’ argument is about islands in fitness landscapes. The ‘sectors’ referred to in the McLaughlin paper are sectors within proteins. You are conflating two utterly distinct concepts.
keiths argues that evolution is not stranded on islands of function because the “landscape” is multi-dimensional.
Even assuming the truth of the premise that the “landscape” is multi-dimensional, it does not follow that evolution is not stranded on islands of function.
keiths argues that if we offer evolution more ways off the island, it follows that evolution will find a way off the island.
Even given more ways off the island, it does not follow that evolution will find a way off the island.
keiths appeals to mathematical abstractions and hypothetical possibilities, but they all run up against the hard facts of reality. The living organism. Actual physical entities. Real genotypes. Actual reproductive rates.
Imagination vs Facts.
DNA_Jock:
Redundancy is good. Anyway, we’ll probably have to repeat it a few more times before Mung gets it.
keiths:
Well gee. I apologize for thinking that your OP was in some way relevant to petrushka’s request. My bad.
Someone else want to start a new thread that’s actually relevant?
keiths,
It’s not as if I did not ask you to clarify the “ID argument” you claimed to be responding to.
If you actually did that and I missed it, I guess that’s my fault. If you didn’t, I think an apology is in order.
Sigh.
Not even one step away, Mung. You’ve got two copies of the exact same genome, Mung, neither of which have undergone any new mutations YET.
The second copy will provide a basis for future mutations to occur without detriment because the original (“good”) copy will still function
Get back to us with your question about “how far away on the landscape” when you have a specific change to the phenotype which is visible to natural selection (or visible for that matter to the machinations of your god Designer).
Sigh. This just isn’t so difficult. It’s ninth-grade biology. You did graduate from middle school, didn’t you?
The McLaughlin paper is relevant, but for reasons completely different from those you cited.
We’ve established that you don’t understand evolution, or ID, or the ‘islands of function’ argument, or my counterargument, or the McLaughlin paper, or even geometry.
What else will it take to persuade you to go off and do the learning you’ll need in order to fashion a coherent argument?
keiths:
Yet another claim made with absolutely no evidence or argument to support it’s veracity. keiths says it, therefore it must be true.
So, keiths, can you please explain the posts by petrushka and Alan Fox, and why you did not explain to them that this thread was not about “islands of function” applied to proteins? Did you just not bother to read what they wrote?
See this.
Mung,
Read the OP. If you still don’t get it, try this.
keiths:
You’re the one conflating the two, not me. In the OP they were two different metaphors. Now, according to you, they are the same. Awaiting your apology.
Mung,
Now you’re just flailing.
Read my comment again, and pay attention this time:
A protein is not a fitness landscape. A fitness landscape is not a protein. An island is not a sector. A sector is not an island.
Think about it.
damitall2 on December 7, 2012 at 9:31 am said:
And keiths said…
Allan Miller on December 7, 2012 at 10:04 am said:
And keiths said…
petrushka on December 7, 2012 at 11:49 am said:
And keiths said…
petrushka on December 7, 2012 at 7:45 pm said:
And keiths said…
Mike Elzinga on December 7, 2012 at 8:12 pm said:
And keiths said…
Allan Miller on December 8, 2012 at 12:06 am said:
And keiths said…
petrushka on December 8, 2012 at 3:34 am said:
And keiths said…
petrushka on December 9, 2012 at 7:15 pm said:
My interest is to respond specifically to gpuccio’s claim that major blocks of code — mainly protein domains — are isolated and unreachable.
And keiths said…
And Mung said,
Enough already.
What hard facts, Mung?
What is one single fact, published in a peer-reviewed scientific journal, which indicates that “actual physical entities, real genotypes, actual reproductive rates” are typically trapped on whatever fitness island they’re currently on?
Facts, Mung, you needz em. You ain’t got em.
Lenski’s got em, and he’s already proved you dead wrong.
hotshoe,
And yet so many people are talking about proteins. And neither keiths, nor you, saw fit to get them back on topic. Go figure.
Let me quote once more, for your benefit:
Now if keiths didn’t write tho OP to address this, then what on earth was the point of the OP? And contrary to your claim, and the assertions of keiths, I don’t appear to be the only one who thought this thread was about something other than “fitness landscapes.”
I’ll await your apology.
Mung:
Mung, the fact that fitness landscapes and proteins are separate concepts doesn’t mean that you can’t use fitness landscapes when talking about protein evolution. Please tell me you understand this.
And I’m very curious, so let me ask again: What is your educational background?
petrushka:
Maybe you can find someone who will start a thread to discuss your specific interest. I tried. Check out Guano.
Two observations:
After a great many posts I have seen no positive evidence for the exixtence of a designer, nor any discussion of the designer’s motives, capabilities or methods. The argument for design remains stranded on the island of missing transitionals. Marooned and isolated from evidence.
And despite a lot of word lawyering, no ID proponent has demonstrated any reason to believe that any process not demonstrated by Lenski is required to enable new functionality. In fact it has been demonstrated that completely new protein functionality is well within reach of Behe’s Edge.
Furthermore, it has been demonstrated that coding sequences are quite tolerant of variation, not unique combinations, not needles in a haystack. Perhaps Mung would like to put his interpretation to the test and contact the authors of the paper he thinks supports isolated islands.
Would it make any difference? You are not here to be constructive.
I suggest you re-read all the threads dealing with dFSCI and then *you* tell *me* how to determine dFSCI.
You sound just like Joe now Mung, you realise that right?
The evidence that would convince me that evolution is guided would be, well, evidence.
Not “prove that it’s not unguided”. Is that so complex, Joe?
What’s the point? Is there nobody who supports ID who can both defend an actual position and who is willing to leave the safe confines of UD?
Mung,
You seem to be trying to make some capital out of distinctions between protein space and the fitness landscape.
The fitness landscape is really about genotypes – in the modern world, raw DNA sequence. Proteins are phenotype – but they are built from genotype. We can map either. Genotypes differentially affect mean reproduction rate of their bearers because they have some effect on phenotype, which causes, by one means or another, more or fewer surviving descendants. Those phenotypes, if they are protein, have different ‘functions’ – biochemical functions – depending on the chemistry of the peptide.
Let’s try and break it down. Say we map all possible 1-base-pair segments of DNA. Easy peasy. A, G, C, T. The space has 4 slots, and you could map it as a simple plane, all points one step away. You may wonder how a single base could affect ‘fitness’, but of course it could, because we are abstracting a particular locus. The 30,000th base on the p arm of chromosome 17 …
It is taken as a given that the mapped genotype is part of a wider genome that possesses the capacity of replication. When we construct a fitness landscape for a segment, we are looking at the effect of all the possible variants of that segment on that basic replication rate.
So we may find that a T as the 30,000th etc etc is lethal, that A, C, G are not, but that they rank in order of their effect on intrinsic reproductive rate. So we can elevate the squares accordingly. T is hopelessly unfit, a black hole, but the squares A, C and G are at different heights depending on the mean fitness of their bearers.
Imagine we could construct a probe and its output was this simple 4-square fitness landscape. Point the probe at base 1 of chromosome 1 and start moving it along. As you do so, the 4 squares may mostly remain flat, but every now and then there will be a blip as one or more possibilities has a greater or lesser effect on the reproductive rate of its bearers.
We could expand this space the ‘other way’, by encompassing more bases. A 2-base space would contain 16 possibilities. But there is no representation that allows those 16 to be placed on a plane with a correct representation of ‘relational’ distance for every element. So the plane becomes even more abstract. Every time we add a base to the set – to widen the tip of our ‘probe’ – we add a dimension. But notionally, we can still flatten that multi-dimensional parameter to the plane, and keep fitness as height.
Now, where protein comes in is that many parts of sequence (though only a few percent) are translated (3-)bit for bit from one sequential form into another. The raw, replicated genotype can specify part of a folded peptide, with biochemical properties. These certainly affect fitness and, for those regions, are the prime reason why our ‘probe’ would detect an undulating landscape instead of a flat one. Not all protein variants are equally fit.
One could, if one wanted, map the peptide sequence directly onto a landscape – Arg-Val-Leu instead of AGAGTATTG. The interesting thing here would be the biochemical properties of the peptide – the ‘function’. This does not automatically tie back to fitness. Both are contextual, but one is within the context of the biochemistry of the wider genome, the other in terms of reproductive rate.
The set of all 99-base DNA sequences maps directly onto the set of all 33-amino-acid peptides. If we stuck our probe, with sensitivity turned to ’99-bases’, onto a segment of a coding region, our output would be the fitness landscape of all possible variants of that peptide at that position. Variants of that position in a real population are likely to be clustered at or near a fitness peak, thanks to NS. The rest of the landscape may indeed be flat – it is entirely possible for a current population to be on the only peak. Does this mean we that we are forced to the conclusion, if we saw that, that ‘function’ was islanded? No. Because we are looking at a snapshot. If a peak is settled, downstream consequences elsewhere in the genome may mean that variation now is detrimental. But in earlier history, the landscape was different, and navigable. And within that landscape, because that is the nature of peptides, there were many different functions. We can demonstrate this empirically by randomly sampling peptide space. It’s stuffed with function.
So there’s another dimension to consider. We don’t just want to run our probe along a genome, we want to keep it fixed on the same segment, then run the tape backwards and and watch as the segment, as part of a wider genome, passes through history to its earliest origins. Only then could we really say whether the function and peak now occupied was at all ‘unlikely’. Likelihood depends on density of hits within the permutation space.
I know we can’t do any of this! It is an exercise in visualisation.
How did this (11.10am) end up here?
And my particular point was that, since the aa-sequence/function relationship in so many proteins is proving so very plastic, there may be a whole host of ways in which a particular peak in a fitness landscape can be reached.
A point which rather devalues the argument-from-huge-numbers so beloved of kairosfocus (which was already a strawman argument anyway); and for the attempted making of which I was banned from UD. Twice.
gpuccio’s last step before inferring X has dFSCI :
“1c) after careful observation and consideration, gpuccio isn’t aware of a ‘necessity mechanism’ that explains X to his satisfaction;”
Come on Mung, there is a 460+ comment thread here devoted to “dFSCI is either circular, or useless, or both”. I believe you ‘participated’.
But do tell, how would *you* determine dFSCI?
The same reason as you when you quote people and leave off the last, critical, sentence.
On the first page of results:
yeah, Mung, you are right, that’s not an argument they make at UD. As long as we create the new term ‘MungRight‘ anyway.
What gpuccio has asserted is that protein domain codes are like lock combinations: they are either correct or incorrect. As some folks at UD have asserted, peaks of fitness have vertical walls. You are at the top or at the bottom of the ocean.
Gpuccio uses this metaphor to argue that protein domains are all or nothing. They could not have evolved by steps.
What McLauhglin has demonstrated is that this version of the landscape metaphor is incorrect.
What I want to ask gpuccio is, given that most modifications of protein elements do not destroy their fold or basic properties, how does a gpuccio designer apply intelligent selection?
Without testing function within a living system that exists within a physical environment and ecosystem? Which of the chemically active sequences will maximize reproductive success in the current context? How does the designer know without trying them out?
I first asked this question a couple years ago, and so far all I have gotten in resopnse is bannations and insults.
The Lenski experiment adds a twist to my question. It appears that even in the context of a living organism there are neutral substitutions. So my question to Mung and gupuccio is how does a designer know which of equivalent variants will become essential in some future context? How does intellient selection work?
In the Lenski experiment it took 20,000 generations for the correct neutral variation to occur. Why does appear, when you look at evolution in pathetic detail, that it works by trying every variation?
Allan, to Mung:
Not even that. As unbelievable as it seems, Mung is actually confusing the “sectors” of the McLaughlin paper (physical structures in a physical protein) with islands in the fitness landscape (abstract structures in an abstract space).
Read Mung’s comment, noting the words and phrases that he emphasizes and the conclusion that he draws:
Mung doesn’t understand the McLaughlin paper, so instead he has merely scanned it for words and phrases that sound “islandy” to him. The distinction between the physical and the abstract gets lost in the resulting muddle.
Poor Mung.
Mung is hung in a Beagle battle puddle muddle?
And if Alan is covertly posting at UD, and the UDenizens continue to cleanse any reference to the identity and attributes of the designer, and continue to take the argument back to Darwin’s warm little pond, then:
Mung is hung by fox in socks in a tidy ID hidey Beagle battle puddle muddle.
Perhaps it would be simpler for Mung just to explain why Mung is an ID supporter at all.
Mung, what is the compelling evidence that causes you to support ID?
I understand you don’t want to go off-topic so perhaps someone will be so kind as to create a new thread just for that topic. Of course, it’s not worth it but finding that out will be worth it all on it’s own.
Given any research paper it’s easy to predict what words will be highlighted when it makes it’s way to the front page of UD.
unexpected
surprising
rethink
overturns old data
yet we never seem to get any close to ID reining supreme.
One of the questions posed by Mung deserves a straight answer, and that question is whether we are willing to consider evidence for ID and whether we would modify our positions if it appears.
I for one am genuinely interested in some questions posed by ID advocates. When I first encountered them I genuinely did not understand the science and could not respond to them. I am always looking for great conundrums. They are entertaining and amusing and they force me to learn stuff.
The questions posed by Douglas Axe and Behe and gpuccio are not bad questions, and they deserve detailed answers. Personally I believe they have been adequately answered by Lensky and Thornton and McLaughlin. Et al.
But I would like to have a genuine dialog regarding these issues, something more interesting than boldfacing isolated words like design that happen to appear in journal articles.
I would like to hear from gpuccio his method for knowing how to produce new features by intelligent selection. I think McLaughlin has adequately restructured this line of argument by demonstrating that raw sequence function is not sparse. So the problem for the designer is how to recognize utility in living context without trial and error.
It would also be interesting to hear why it apparently takes the designer an average of two million years to make a new protein domain. Since McLaughlin has demonstrated that transitional proteins are fairly common, and completely new ones can be created by only two mutations, it calls into question the exponent that gpuccio is using to calculate dFSCI.
I’ve actually waited for years for gpuccio to explain why a sequence cannot be isolated from its cousins. I’ve offered the analogy of the Basque language, and I’ve shown sequence histories from simulations that show how all elements of a sequence can be incrementally altered until cousin sequences are completely isolated from each other.
I am certainly interested in evidence for design. But it will need to be real evidence, not merely a purported problem for evolutionary accounts.
Mung, quoting vjtorley at UD:
Mung’s commentary […]
This risks drifting somewhat OT, but for the record adaptations by their very nature can only be built by NS (outwith ‘poofing’ Designers. Can you think of a third possibility that a proportion of TSZ-ers should espouse, just to avoid accusations of groupthink?). This is not only the prevalent view in TSZ, it is the prevalent view in evolutionary theory, and arises almost by definition.
An adaptation enhances the reproductive rate of its bearers over that of those who do not bear it. That’s Natural Selection. If it’s just an ornament with no use, making no difference to its bearers’ reproductive success, it’s hardly an adaptation.
Complex adaptations … complexity does not arise as a result of NS, and I’m not aware of anyone at TSZ who thinks that it does. You don’t leave more offspring simply because you are complex. Many organisms leave more offspring as a result of becoming more simple – see parasites, for example.
What’s with the ordering of non-reply comments?
It looks as if I move a comment to Guano, and there were replies to that comment that I didn’t move. That’s probably what causes the strange order.
Yes, I reversed one such move, and it has changed the order. I’ll try to fix others later. Go an errand to do first.
The relationship of NS and complexity is further discussed here at Sandwalk and on to the linked Pharyngula piece. Arguments about the relative strengths of the various evolutionary ‘forces’ (NS, drift, mutation, recombination, mixing) rumble on from time to time, but no informed commentator, however caricatured as “panselectionist” or whatever its ‘random’ antithesis might be, denies that the ones they aren’t focusing on are important. And the relationship adaptation <-> selection is pretty much settled.
[eta: which is not to say that every last step in the building of a complex trait we view as an adaptive whole must have been incrementally beneficial.]
All such disagreements look like chicken and egg disagreements to me. What seems to be missing from the argument over selection is the rather stunning fact that neutral and nearly neutral variants exist. That’s kind of the topic of this thread.
The mere fact that selection is not so tight as to prohibit variation, and biochemistry is not so sparse as to prevent functional variants is why evolution works.
keiths,
Yes, I managed to scan all the way through the first paragraph. You need to make up your mind whether we’re going to discuss the paper or not. You don’t get to have it both ways.
As far as confusing the physical and the abstract, that’s precisely what you have done with your “fitness landscape” of trillions and trillions of dimensions. It’s totally divorced from the physical reality of the living organisms.
And you really do seem to be confused about what IDers say. Here’s a quote from a recent post by KF at UD:
IDers apply the ‘islands of function’ metaphor to proteins. If you spent more time quoting IDers and less time making things up we’d be past this point already.
keiths,
How does ‘islandy’ sound. I’ll try to avoid the usage of ‘islandy’ sounds in the future. But then, your whole OP sounds ‘islandy.’
So, if you’re to be taken seriously, you want me to discuss ‘islands of function’ without any words that sound ‘islandy.’ Right.
Mung:
No, what I want is for you to understand and discuss the actual ‘islands of function’ concept, instead of your bizarre misinterpretations of it. Use “islandy” words, by all means, but first understand what you are talking about.
What you’re doing is scanning the McLaughlin paper for “islandy” words and phrases instead of understanding the concepts. You’re not fooling anyone.
Mung:
Of course they do. I’ve never stated otherwise. What I’ve been telling you is that proteins and fitness landscapes are not the same thing.
Slow down and think about it, instead of rushing headlong into another foolish mistake.
I addressed this issue two days ago:
Mung, quoting KF
That is nonsense. It is like saying that half a dozen surviving photographs from someone’s life support the view that that individual changed in huge leaps, or was refashioned afresh from the dust of the earth between snaps.
Say all the possessors of a particular ‘body plan’ really do descend from the same ancestral species. If that species possessed some character that would be a useful ‘missing link’ in a body plan series, what factors do you think might impinge on someone being able to locate its remains a hundred million years hence? How widespread, persistent and likely to be found can we expect a single species‘s remains to be over that time period?
A successful body plan founding a dynasty will speciate and grow in number, and will, as time goes on, be found in more and more accessible, and less and less decayed, strata. All factors that would make the appearance in the fossil record seem sudden and miraculous.
I prefer the description of a ‘fitness landscape’ given by Douglas Axe to the one offered in the OP:
We can dispense with all the talk of undefined and unknowable other dimensions as superfluous.
I’m not defending it (yet). I’m pointing out that keiths doesn’t appear to understand the arguments he’s presuming to criticize.