Information and Genetics

I considered opening a similar thread and probably should have. I suggest the information theory discussion on the "Give your one best shot - against evolution" thread be moved here. I'll post a note over there.--Percy

These are replies to Message 220 and Message 221 from SLPx over at the Give your one best shot - against evolution thread.About the Kimura paper, sure, send me a copy if you want: percipient@. But all I'm really seeking is an explanation for how natural selection could create new information. As I said originally, my only guess is that it might involve equating new genetic information with recombining existing alleles permutationally, but it would nice to know what Kimura was actually thinking.About Tom Schneider's assessment of Fred's website, is there a link?--Percy

Fred writes:

---

Percy, we too are at a dead end. If you insist that a computer is not an intelligent device, or represents intelligence, then what is the point in debating further?

---

Whether a computer is intelligent or not is a side issue, but there's no point playing musical definitions. Obviously I was using definition 1.You claim that information can only be sent and received by intelligence, but even by your definition 5 it takes no intelligence to send or receive information. The starlight we see is full of information, but it was not sent by intelligence. And you can attach a message to an arrow and shoot it across a battlefield, but the tree it strikes is not intelligent.

---

Like Mark, you have concocted a version of information that is not falsifiable.

---

If we refer to the Shannon paper you're so fond of referencing, right on page one he says:

Frequently the messages have meaning; that is they refer to or are correlated according to some system with certain physical or conceptual entities. These semantic aspects of communication are irrelevant to the engineering problem.

In other words, the interpretations that intelligence builds around the bits flitting through cyberspace are irrelevant to information theory. And if you look at Figure 1 you'll see that he defines a communication system with an information source, a transmitter, a source of noise, a receiver and a destination. There is no intelligent sender or receiver, and the word intelligence does not appear anywhere in the paper.

Fred writes:

---

I have even produced evolutionists who are on the record that GAs are bogus examples of evolution, you only have your opinion to support your claim.

---

But this is the fallacy of argument from authority, plus I lead a sheltered life and have never heard of your evolutionists and so have no idea what they have said. Can you reproduce the relevant arguments here?

---

Your simulation does not reflect reality. It employs severe truncation selection, yet there is no evidence such selection occurs in nature, and most evolutionists reject such a wild notion.

---

My model reflects reality, it's only a matter of degree. The selection criteria can be modified to be whatever you desire, Fred. The model was only intended to falsify your claim that random mutation cannot produce new information, which it does.Let's take a different example, say of a physics equation developed by randomly placing letters into a template, say, a=bc², with the selection criteria being detailed studies of the equation's correspondence with reality. Is it possible that E=mc² could pop out? Of course it is. In fact, given time (mainly for applying the selection criteria, since a computer could arrive at E=mc² in an eyeblink), it's inevitable. Randomness can generate information, and with the right selection criteria it will be appropriate to the task at hand.This is precisely what happens with evolution. The total environment places selection pressures upon organisms. Those that survive to reproduce on average possessed some advantage, and the genes reflecting this advantage are passed on to offspring with modification, both from simple permutational recombination of the haploid egg and sperm chromosomes and from mutation. The offspring are in effect a set of experiments to see which will survive to repeat the process. Mutations that confer an advantage will be selected for. This process is consistent with information theory.--Percy

Fred writes:

---

See my response to Page on tree rings.

---

Yes, I know, tree rings aren't information because they aren't a code. This argument is as wrong for tree rings as it is for starlight. Starlight is information. It is uncoded information, or at least is not represented in a form we would normally think of as a code, but it *is* information.Chinese writing, alphabets and binary bytes are merely codes that humans have found convenient for representing information. A code is a way of recording, representing and transmitting information, but is not itself information (unless you're interested in the code itself). Information can be coded, and uncoded information is still information.What's more, it takes only a modest change in perspective to see that light *is* a code. The spectrum of light is represented as waves of different wavelengths, intensities and polarizations. It's a code that can be partially decoded by our eyes, but which really begins to give up its secrets when subjected to spectrum analyzers and so forth.

---

As you somewhat alluded to earlier, when Shannon wrote his paper he was not inventing information theory.

---

You must be thinking of someone else.

---

His purpose was to establish mathematically communication throughput and efficiency. From his paper developed the modern concept of information as understood by today’s info theorists. It was Shannon’s equivocation that is today viewed as Shannon information.

---

Shannon's equivocation has to do with the randomness and uncertainty that I addressed in earlier messages. But the point I was making this time is that the semantic aspects of information are, as Shannon points out, "irrelevant to the engineering problem." Semantic interpretations, which derive from intelligence, are irrelevant to information theory.

Percy writes:

---

But this is the fallacy of argument from authority...

---

Fred replies:

---

Not if the person has expertise in the field.

---

Huh? It would be a fallacy even if you were making claims about relativity while citing Einstein. Maybe you haven't heard of the fallacies of debate, but citing authority is one of them. It is used as a means of avoiding having to support a position, eg, "Famous scientist X agrees with me, and that's that."If you'd like to provide some discussion supporting the positions of the evolutionists you mentioned I'd be glad to look it over.

---

BINGO! Chance of success = UNITY. NO INFORMATION GAINED! Find me one single info theorist in the world who agrees with you that an experiment guaranteed to reach a pre-determined target has produced new information without a sender.

---

Once again you're introducing requirements unrelated to information theory, and if I were to do as you suggest and cite info theorists who agree with me it would once again be the fallacy of appeal to authority. A debate is won by supporting your position with evidence and argument, not by saying, "Genius X says so, nyah, nyah."About the outcome being predetermined, just run the experiment in 1900 before Einstein came up with E=mc². Or, since E=mc² was chosen just to have something concrete to focus on, imagine we're only considering unknown equations and apply the afore described selection criteria.

---

We are at a dead end, Percy. You are flat wrong.

---

You may be at a dead end, but I'm not. It's fairly clear where your misunderstandings of information theory lie, and each time you restate them the path to correcting you is pretty clear.

Percy writes:

---

Those that survive to reproduce on average possessed some advantage...

---

Fred replies:

---

This is a textbook tautology, rendering the statement useless.

---

Then so is, "The fastest car wins the race."I guess calling it a tautology is what you do when you're otherwise at a loss, because as an objection this makes no sense. That was just one sentence in a description outlining the process for which I created the C++ model. It's the part that describes selection, where the closer the sequence becomes to the environmentally imposed sequence, the more likely the offspring is to survive to reproduce.

---

Incorrect. Even when evolutionists assume strong selection, they admit that it has at best a 1 in 50 chance of fixation.

---

Once again you're citing unknown evolutionists saying things that heaven only knows were really said, and obviously I wasn't trying to describe things at the level of detail of fixing genes within populations.

---

No it isn’t, because information cannot arise naturalistically. There are no known examples in the universe to counter this law of nature.

---

And yet descent with modification through selection and mutation happens. And since we understand reproduction and mutation we are able to model them, just like any other natural process we understand.Two different threads of objection to your information theory position are being followed in this thread, and both are valid. One provides illustration that the very process that you claim information theory renders impossible most definitely has occurred and is still occurring. The other shows the ways in which you misunderstand information theory, for instance by attaching unrelated requirements like utility and intelligence.--Percy

I was only attempting to repeat Creationist information theory as explained to me by Creationists, for instance, what Fred was saying in this thread. Can you make this sound like more than just an arbitrary list of properties? For example, where is the imperative that a code be "specified and complex?" Why must it have a "representational function?" Why do you exclude the "inherent physical properties of the system" as a factor in determining the code?The primary problem you face is that Gitt has invented a theory of information purporting to demonstrate that the process of evolutionary innovation is impossible. But we can demonstrate evolution innovation in the lab, and when theory conflicts with evidence, guess which loses?Just as we can model other natural processes such as the weather and planetary orbits using computers, so can we do with the process of random mutation and selection. See my C++ program that simulates ring counter evolution. A description of the program can be found at Message 142 of the Give your one best shot - against evolution thread.--Percy

Hi Fred Seigneur, welcome aboard!I'm afraid I agree with John, though I would express it a bit differently. Such an effort as you describe could not fail to yield many interesting Bible quotes. As the many rebuttals to The Bible Code have amply demonstrated, and as anyone with an appreciation of statistics already knows, data mining lengthy texts (in this case a genome somehow translated to text) will always produce many matches, and some will probably match passages from the Bible. But the same exercise would also find passages from the Koran, the Bhagavad Gita, Das Kapital, Catcher in the Rye and The Cat in the Hat. This webpage uses the text of The Bible Code II to provide an example of just how such data mining works to extract supposedly amazing quotes. It's good fun and a fine example of statistical effects, but is otherwise meaningless.--Percy

Mammalian-identifying features, chosen for their ability to distinguish between other classes of organisms, are not arbitrary.What I actually asked was why the requirements for your code appear to be arbitrary. To repeat, where is the imperative that a code be "specified and complex?" Why must it have a "representational function?" Why do you exclude the "inherent physical properties of the system" as a factor in determining the code? You haven't defined A, B and C - I don't know what you're saying. You haven't described the "type of code Gitt describes," so I'm unable to evaluate your argument.--Percy

Hi Dillan,Thanks for making the effort to clarify through quoting, but your words and my words appear at the same quote level. If I hadn't recognized my own words I would have thought you wrote all the quoted words and that, since it begins with "dillan writes," you were replying to yourself. There's a preview button that helps figure out how a message will look before you post. Covariance is a statistical measure that is probably not the most accurate term to use in defining the term "code". Covariance analysis would probably be very useful in studying a code with which you weren't already familiar, code breaking, for example. As if you've conducted a poll. Tree rings are most certainly a code. You're making the same claims but still not explaining anything. Why does a Gitt code require specified complexity, and why do you think a genome contains specified complexity. What is representational function, and why is it required of a Gitt code? Some "information systems" are the product of intelligence, some aren't. Quite obviously, since we can observe this in the lab, the process of mutation and selection takes place with no intelligent intervention whatsoever, including changes in complexity (as measured by amount of Shannon information) in both the positive and negative direction.You're seeking a set of criteria that when satisfied leads inevitably to the conclusion of intelligent design, but what you've described here is simply an attempt to ignore real world data through abstraction. And it isn't even a mathematical abstraction. Shannon information is very mathematical. The information carrying capacity of a channel in the presence in the noise can be expressed quantitatively. What you're describing is qualitative because your bottom line argument is based upon subjective assessements of patterns rather than quantitative measures.Your classification of codes into domains A, B and C appear arbitrary to me. For example, tree rings code for how good the growth season was how many years ago, a clear semantic. Starlight codes for huge amounts of information that fills volumes of books on cosmology, and is still growing, more very clear semantics. The definitions of your domains seem constructed from a premeditated desire to characterize genomes as intelligently designed rather than from any objective criteria. Not according to any scientifically accepted and demonstrated definition of information. What you're really talking about isn't mathematical information but knowledge or semantic meaning, another creature altogether. The knowledge (as contrasted to mathematical information) about the natural world recorded in science books and journals would have existed whether or not we were here to take note. Our mere existence did not suddenly bring knowledge into being.--Percy

This discussion represents an opportunity to explore and understand these issues for ourselves, something the forum guidelines try to encourage. Why cut ourselves off from this learning opportunity by instead engaging in the fallacy of appeal to authority? The best way to make your point in a discussion is by informing yourself so you can argue from a position of knowledge. I could as easily reply, "You are the one making the claim that tree rings are not a code, so the burden of proof is up to you," and where would that get us. I have argued that things like tree rings, starlight and genomes can be deciphered by scientists to provide us a wealth of information. How is deciphering tree rings to determine age, growth and climate patterns any different than deciphering genomes to determine patterns of inheritance?You quote Gitt: This merely repeats your claim. Why is this a requirement of a Gitt code? In other words, what is Gitt's justification for this requirement? Even further, since a genome has no behavior not explainable by physics or chemistry, Gitt's own definition excludes it as a Gitt code.You follow this quote by saying, "Read my links." I'm going to be honest with you. I'm not going to read your links. I think you should make your own points in your own words and only cite links and references as support, not primary, material. It is not your responsibility to educate me about Gitt, but I am not debating Gitt. I am challenging your position that genomes must be the product of intelligent design, and you are defending it. I think you're confusing Gitt with Dembski. "Specified complexity" is Dembski's term, and he believes the genome has been specified by an intelligent being. But you can't simply declare that a genome has specified complexity. No one would deny that genomes are complex, especially in the decoding, but you have offered no evidence that they are specified. This is also untrue, because there is no requirement that an information system specified by an intelligence be complex. Assuming I qualify as intelligent (often questioned, but grant me the benefit of the doubt for this argument), I've designed plenty of simple information systems. This website is an example.You quote Gitt saying this: When Gitt says "information" he is actually talking about human encoded information, and so he is begging the question. A scientist's record of tree ring widths in his notebook is human encoded information, but the tree rings themselves are also encoded information, just not by a human. And yet you can't point to any scientific theory of the origin of life which relies upon the impossible. Actually, I would argue that your assertions can either be supported by evidence or not. So far I have seen no evidence supporting your views. While that does not mean they aren't true, it certainly isn't very encouraging about the possibility. Genes also deal only with the "inherent physical properties" of matter. The DNA sequence does nothing more than guide the behavior of other matter with "inherent physical properties." And possession of this quality does nothing to exclude possession of a semantic associated with the code, as I already described for tree rings and starlight. As does everything else, including genomes, meaning that we're comparing apples and apples, and this still leaves you short an explanation for why this excludes the possibility of semantic content. We read the information encoded in starlight and learn whether the star is old or young, large or small, near or far. I think when you say "encoded" that what you really mean is "encoded by an intelligent being." By this definition I agree that tree rings and starlight are not encoded by intelligent beings, but then neither are genomes, at least not that you have any evidence for. I would refer back to the point I made earlier. Links and other types of references are best used only to support, not make, arguments. I'd be glad to debate Gitt if he'd like to join us here, but Gitt's not here right now and I'm debating you.--Percy

I think Paul is referring to his Message 110 from the Proof against evolution thread in the Origin of Life forum. Here's the link to the AiG article: As Gitt accurately says in this article, with Shannon information, "whether a chain of symbols has a meaning is not taken into consideration." Shannon says almost precisely the same thing on page one of his landmark paper: "Frequently the messages have meaning; that is they refer to or are correlated according to some system with certain physical or conceptual entities. These semantic aspects of communication are irrelevant to the engineering problem."In other words, on this point Shannon and Gitt agree.But Gitt goes on to define five distinct levels of information. I enumerate them here - the lowest level is the familiar Shannon information:

1. Statistics. In other words, Shannon information.
   ​
2. Syntax. These are rules for how the symbols of a code may be combined. Why he excludes Shannon information from this level isn't clear, since Shannon information uses codes with syntax. For example, Hamming codes restrict the way that 0's and 1's can be combined in a message so as to permit error detection and correction. The Gray Code permits progressions of code symbol sequences where only one symbol changes at a time, which is useful in self clocking systems and in reducing power requirements by minimizing the toggling of signal values in cascading circuits. So to repeat once more, Shannon information definitely uses codes that possess syntax.
   ​
3. Semantics. In other words, meaning. Shannon information definitely does not include meaning. While we agree that messages have meaning, the meaning of the message is irrelevant to Shannon information, while it is apparently a factor with Gitt information.
   ​
   This factor introduces all the complexities of language. Indeed, Gitt even relies upon the example of language to introduce his definition of semantics. However, not only are there no quantitative measures of semantics, the semantic meaning of a message is indefinite, non-distinct, fluid, inconstant, etc, etc, etc. Puns and double entendres are simple examples of the indefiniteness of semantics. We experience this indefiniteness all the time when we speak to other people - clarifying what we just said so it may be accurately understood is a large part of many conversations.
   ​
4. Pragmatics. As near as I can gather, this is the intent of the message, in other words, the reason you sent it. This would seem to possess as much indefiniteness as semantics, and is probably actually just one aspect of semantics rather than a higher level.
   ​
5. Apobetics. You gotta love made up words. I couldn't find a single use of the word outside the context of the Creation/evolution debate, it isn't in my desk dictionary, and it isn't at the Meriam Webster site. Gitt defines apobetics as purpose. How the purpose of the message is different from the reason you sent the message, which is the previous level of pragmatics, is not clear, and they both would seem to be part of semantics. I really see only three levels.

It was interesting to see this summary of Gitt information, especially since it's by Gitt himself, but the logic which flows from these definitions by which Gitt concludes DNA is the result of intelligence is severely flawed. It reduces to simply claiming that DNA has a semantics, a purpose, an intent. I think we would all grant that it has semantics, but the purpose and intent that Creationists claim to see are simply enumerations of what DNA does. And they disallow all claims of purpose and intent for other codes like tree rings and starlight because they only accept purposes and intents that are related to important things like heredity and reproduction, but not for less important things like how old is a tree is or how large is a star. Naturally what is important and what isn't are value judgments they themselves impose.--Percy [Accidently posted before cut-n-pasting in the final edit. --Percy][This message has been edited by Percipient, 09-06-2003]

I'm looking at Gitt's definitions right now and I think this is a quibble about details of wording. The two points I made relevant to this discussion were:

1. Pragmatics and apobetics (a made up word not adopted by the fields of either information theory or semantics) are just sub-parts of semantics. There are really only three levels: statistics (part of Shannon information), syntax (also part of Shannon information) and semantics (specifically excluded from Shannon information).
   ​
2. Semantics is indefinite, unquantifiable, ambiguous, relative to the interpreter, etc, etc, etc.

DNA only obeys its "inherent physical properties". There's nothing there but matter, physics and chemistry. Any "representational function" you perceive is a semantic superimposed on DNA by yourself, just as it is for tree rings and starlight, and is not an inherent property of DNA. This is because, as I characterize above, semantics is an interpretation. It is inherently ambiguous. It is why reviews of the same book or movie differ so dramatically. Gypsies interpret tea leaves as having a semantic that foretells the future, as do astrologers of the planets. The semantics come from them, not from the objects they study. You're simply playing tea leaves with the concept of DNA. As Mammuthus and Mark have both said, and I agree, you need an approach that is objective and scientific, not one only acceptable to conservative Christians of a scientific bent.You're approach is also not all that dissimilar from those who argue for intelligent design because the earth is at exactly the right distance from the sun for life, or because the universe has precisely the properties necessary for life.We haven't touched on another of the major problems for ID, and that comes from within the very religious community from which it springs. Evangelical Christians require an intelligent designer who is actually the Christian God of the Bible. As long as that requirement is fulfilled they don't particularly need the ID'er to be called God in secular settings such as public school science classrooms. But what the ID movement is instead providing is an intelligent designer of a universe billions of years old where there was no six days of creation, no garden of Eden, no original sin, no flood. ID is doomed to be disowned by its own parent. The superficial reason that ID failed recently in Ohio is because of opposition within the educational and scientific communities, but the underlying and more substantive reason is lack of support from within the conservative Christian community.Speaking for myself, but I'm sure I'm not alone among evolutionists, I sincerely wish that God had left behind evidence of his existence. I do not reject Gitt not because I'm philosophically opposed to the idea of God. I definitely am not. I've stated many times here that I am a sincere believer in God. But I'm not a sucker to be taken in by crying statues and virgin images, and neither are most other sincerely religious people. Just because you believe in God doesn't mean you believed it every time (or any time) Oral Robers, Jimmy Swaggert and Jim Bakker claimed that God had appeared to them in a vision last night (listing those names has probably dated me - plug in the names of the currently popular TV evangelists). And so thinking Christians have to be careful because they're almost literally surrounded by those trying to take advantage of their credulity, and this includes the pseudo-scientific gibberish of Gitt's information argument.--Percy

Dillan may be having the same problem I did in tracking down the link you're referring to. Check my Message 130 to see if I got it right. If not, you might want to provide Dillan a link to the message where you provided the link and made the points, because since this discussion has hopped across several threads it isn't obvious where it is.--Percy

In an earlier message I said this: If you go to PaulK's message you'll see where near the end he quotes Gitt saying that semantic information can't be approached mechanistically, though I think Gitt's phraseology is open to other interpretations.I don't have time right now to read all of your lengthy reply to me in Message 144, but in the first paragraph you say this: I'd be glad to help make your job easier, but you'll have to tell me what part of my message you're referring to.--Percy

I quoted this as part of a longer excerpt in my previous message, but I didn't comment on this particular part, so I wanted to respond to it now to let you know that I'm not in a hurry for you to reply to me. In fact, I'm surprised that you respond as rapidly as you do. There's no urgency, so if you're feeling overloaded then please take your time. Agreed. And here is where we part company. If you read Gitt's webpaper and look under pragmatics you'll see where he talks about intent: If pragmatics is defined as the intent of the message, then I can agree that pragmatics is a part of the semantics of the message. But Gitt goes on to become more specific and describe it in the same terms you use above, and I don't agree that the actions of the receiver are part of the information of the message. Let me say it again, just to be clear. I can agree that the intent of the message is part of the information of the message, but not that the actions performed by the receiver in response to the message are part of the information of the message. It is nonsensical positions like this (among many) that caused me to earlier refer to Gitt's position on information as gibberish. Not according to Gitt: If we go by my understanding, then I don't see a whole lot to distinguish between intent (pragmatics) and purpose (apobetics), and they are both part of semantics. And if we go by your understanding then we are left with a weird definition of information where the information content of a message is a function of what some future receiver of that message happens to do in response to it, and if one of the receiver's responses is to send messages to others, which is almost always the case, then this process never ceases with the nonsensical conclusion that the entire universe is a single message.Gitt's ideas include such gibberish because they haven't been submitted to peer review, but only to completely uncritical Creationists hungry to latch onto any likely lookly evidence for God. There are no Creationist scientists working on the same ideas as Gitt who are competing with him for status within the scientific community by introducing their own original contributions. Instead, Gitt's ideas are being submitted directly to uncritical Creationists who then reiterate the ideas in books, on websites and at discussion boards when they definitely aren't ready for prime time. That Gitt's information ideas are mentioned nowhere outside Creation/evolution circles should tell you something. If they had merit then they'd be mentioned in information science and semantic venues, both of which are active areas of scientific research. That they're not says that at this point in time the ideas have not been found to have any scientific merit, and so remain a object of interest only to a particular religious branch.The main reason Gitt engages in this gibberish of ideas is because he wants to lump together both the information and any interactions of the information with its environment, and simply because DNA does this more obviously than other things, making seem more reasonable the flawed argument that DNA could only have been intelligently created.I've been wrestling with myself about why this discussion is becoming so extended. I think the issue still comes down to one you haven't yet addressed, the ambiguity of semantics. I have described semantics several times now as ambiguous, indefinite, relative to the interpreter, etc. I think the examples of astrologers and gypsy tea-readers are applicable, for they are imposing their own semantics upon the planets and tea leaves just as you are imposing your own semantics upon DNA. To argue that DNA has semantics that can only have originated with an intelligence is true, but only because those semantics originate with people like yourself and Gitt. They are semantics imposed post facto and are not an inherent property of DNA.I'm not replying to the rest of your message, not because I'm ignoring it, but because this post is already too long, and because the discussion has moved forward by a couple dozen messages before I found time to reply.--Percy

Hi Fred,I think Rei's response already covers the issues you raise pretty well, so I thought I'd just nibble around the edges a bit and comment on a few of the other things you say in passing. I don't know about the "1 of every 1000" part, but I get your point and it is probably true. But of what possible relevance to the validity of genetic algorithms (GA's) is their current utilization rate in engineering projects? Once again, of what possible relevance are these comments to the validity of GA's? I would think it would be more productive to stay focused on the specifics of GA's rather than upon subjective ruminations about the people working on it. When Rei says that GA's are not trial and error I think he was trying to say that they are not *merely* trial and error. Certainly each mutation is a "trial", and each entity not selected is an "error", but just saying trial and error leaves an awful lot out, since the mutation algorithm and selection criteria can be highly complex and difficult to design and implement. I think it might also be argued that GA's also have some similarity to successive approximation approaches.I strongly disagree with you that GA's have limited usefulness. In fact, as computer power becomes more and more ubiquitous the ability of GA's to provide solutions to complex problems will become of greater and greater value. Perhaps you've heard of Blondie24, the commercial Checkers with an Attitude program developed using GA's. (I wonder why none of the opponents are male? ) It was also described in a recent article in the Communications of the ACM, or maybe it was IEEE's Computer magazine. Anyway, the point is that GA's not only have definite utility, but they're out of the lab and developing commercial products, as Rei also pointed out to you. It's true that very little information was provided about Engeneous, but stating that "it clearly was a trial&error program that I'm sure uses 100% pure truncation selection" completely ignores that information (Forbidden): Further, about your assumption that it used pure truncation selection, first, it is irrelevant. If it helps the engineers improve the design, then it makes no difference what type of selection was used.Second, you have no information upon which to base this assumption. Maybe it used truncation selection, easily possible since this is the simplest approach, and maybe it didn't. The article simply doesn't say.Third, your statement that other types of selection wouldn't make sense is baseless. For example, rather than only selecting the n best in each generation, it might select all that surpass the performance of the previous generation by at least an amount x, with the simulation halting when no member of a new generation accomplishes this. GA's don't even have to take a truncation approach. Instead of generating m new approaches and taking the n best, which is truncation, it might instead just keep generating new approaches until it has n that exceed the previous generations average performance by x, and giving up when the computational time for a new generation exceeds some set limit. There are plenty of selection approaches that make sense beyond simple truncation. First, challenging GA's based on the type of selection criteria used has no validity. GA's are simply software, and the selection criteria can be modified to fit requirements. You specify the type of selection criteria you want, and I'll write a GA program that uses it. From Rei's job description he could obviously do the same thing for you.Second, truncation selection most certainly occurs in nature. For example, in many ecological niches offspring produced in excess of the environment's ability to sustain them die, ie, get truncated. I applaud Rei's not responding in kind.--Percy