|
Author
|
Topic: What Sort of Property is Specified Complexity?
|
William A. Dembski
Member
Member # 7
|
posted 29. August 2002 21:26
What Sort of Property is Specified Complexity?
By William A. Dembski
NOTE: The following essay is a chapter for a book I'm writing. As usual, I hammer away at the bacterial flagellum. Critics of mine may wonder when I'm going to let it go. I'll let go once my critics admit that it represents an insoluble problem on naturalistic terms. Also, I say in this essay that there is no evidence for an indirect Darwinian pathway to the bacterial flagellum. Sorry, but the type III secretory system doesn't cut it. In fact, Milt Saier's work at UCSD suggests that the type III secretory system, if anything, evolved from the flagellum. But even if it could be shown that the type III system predated the flagellum, it would at best represent one possible step in the indirect Darwinian evolution of the bacterial flagellum. To claim otherwise is like saying we can travel by foot from Los Angeles to Tokyo because we've discovered the Hawaiian Islands. Evolutionary biology needs to do better than that. At any rate, the aim of this essay is not to rehash the flagellum, but to come to terms with what sort of property specified complexity is. I hope this essay stimulates discussion on that question.
Specified complexity is a property that things can possess or fail to possess. Yet in what sense is specified complexity a property? Properties come in different varieties. There are objective properties that obtain irrespective of who attributes them. Water is such a property. There are also subjective properties that depend crucially on who attributes them. Beauty is such a property. To be sure, beauty may not be entirely in the eye of the beholder (there may be objective aspects to it). But beauty cannot make do without the eye of some beholder.
The distinction between objective and subjective properties has a long tradition in philosophy. With Descartes, that distinction became important also in science. Descartes made this distinction in terms of primary and secondary qualities. For Descartes material objects had one primary quality, namely, extension. The other properties of matter, its color or texture for instance, were secondary qualities and simply described the effect that matter, due to the various ways it was configured or extended, had on us as perceivers. Descartes's distinction of primary and secondary qualities has required some updating in light of modern physics. Color, for instance, is nowadays treated as the wave length of electromagnetic radiation and regarded as a primary quality (though the subjective experience of color is still regarded as a secondary quality). Even so, the idea that some properties are primary or objective and others are secondary or subjective remains with us, especially in the sciences.
The worry, then, is that specified complexity may be entirely a subjective property, with no way of grasping nature at its ontological joints and thus no way of providing science with a valid tool for inquiry. This worry, though misplaced, needs to be addressed. The first thing we need to see is that the objective-subjective distinction is not as neat and dichotomous as we might at first think. Consider the following three properties: X is water, X is married, and X is beautiful (the "X" here denotes a place-holder to which the properties apply). X is water, as already noted, is objective. Anybody around the world can take a sample of some item in question, subject it to a chemical test, and determine whether its composition is that of water (i.e., H2O). On the other hand, X is beautiful seems thoroughly subjective. Even if objective standards of beauty reside in the mind of God or in a Platonic heaven, in practice people differ drastically in their assessments of beauty. Indeed, no single object is universally admired as beautiful. If specified complexity is subjective in the same way that beauty is, then specified complexity cannot be a useful property for science.
But what about X is married? It certainly is an objective fact about the world whether you or I are married. And yet there is an irreducibly subjective element to this property as well: Unlike water, which is simply part of nature and does not depend for its existence on human subjects, marriage is a social institution that depends intimately for its existence on human subjects. Whereas water is purely objective and beauty purely subjective, marriage is at once objective and subjective. This confluence of objectivity and subjectivity for social realities like money, marriage, and mortgages is the topic of John Searle's The Construction of Social Reality. Social realities are objective in the sense that they command intersubjective agreement and express facts (rather than mere opinions) about the social world we inhabit. But they exist within a social matrix, which in turn presupposes subjects and therefore entails subjectivity.
Searle therefore supplements the objective-subjective distinction with an ontological-epistemic distinction. Accordingly, water is ontologically objective -- it depends on the ontological state of nature and is irrespective of humans or other subjects. Alternatively, beauty is epistemically subjective -- it depends on the epistemic state of humans or other subjects, and its assessment is free to vary from subject to subject. Properties reflecting social realities like money, marriage, and mortgages, on the other hand, are ontologically subjective but epistemically objective. Thus marriage is ontologically subjective in that it depends on the social conventions of human subjects. At the same time, marriage is epistemically objective -- any dispute about somebody being married can be objectively settled on the basis of those social conventions.
How do Searle's categories apply to specified complexity? They apply in two parts, corresponding to the two parts that make up specified complexity. Specified complexity involves a specification, which is a pattern that is conditionally independent of some observed outcome. Specified complexity also involves a measure of complexity, which calculates the improbability of the event associated with that pattern. Think of an arrow landing in a target. The target is an independently given pattern and therefore a specification. But the target also represents an event, namely, the arrow landing in the target, and that event has a certain probability.
Specifications, by being conditionally independent of the outcomes they describe, are, within Searle's scheme, epistemically objective. Moreover, once a specification is given and the event it represents is identified, the probability of that outcome is ontologically objective. Consider, for instance, a quantum mechanical experiment in which polarized light is sent through a polaroid filter whose angel of polarization is at forty-five degrees with that of the light. Imagine that the light is sent through the filter photon by photon. According to quantum mechanics, the probability of any photon getting through the filter is 50 percent and each photon's probability of getting through is probabilistically independent of the others. This quantum mechanical experiment therefore models the flipping of a fair coin (heads = photon passes through the filter; tails = photon doesn't pass through the filter), though without the possibility of any underlying determinism undermining the randomness (assuming quantum mechanics delivers true randomness).
Suppose now that we represent a photon passing through the filter with a "1" and a photon not passing through the filter with a "0." Consider the specification 11011101111101111111..., namely, the sequence of prime numbers in unary notation (successive 1s separated by a 0 represent each number in sequence). For definiteness let's consider the prime numbers between 2 and 101. This representation of prime numbers is ontologically subjective in the sense that it depend on human subjects who know about arithmetic (and specifically about prime numbers and unary notation). It is also epistemically objective inasmuch as arithmetic is a universal aspect of rationality. Moreover, once this specification of primes is in place, the precise probability of a sequence of photons passing through the filter and matching it is ontologically objective. Indeed, that probability will depend solely on the inherent physical properties of photons and polaroid filters. Specified complexity therefore is at once epistemically objective (on the specification side) and ontologically objective (on the complexity side once a specification is in hand).
Specified complexity therefore avoids the charge of epistemic subjectivity, which, if true, would relegate specified complexity to the whim, taste, or opinion of subjects. Yet specified complexity does not merely avoid this charge. More positively, it also displays certain positive virtues of objectivity: specifications are epistemically objective and measures of complexity based on those specifications are ontologically objective. Is this enough to justify specified complexity as a legitimate tool for science? To answer this question, let's consider what could go awry with specified complexity to prevent it from functioning as a legitimate tool within science.
Specifications are not the problem. True, specifications, though epistemically objective are not ontologically objective. The failure of specifications to be ontologically objective, however, does not prevent them from playing a legitimate role in the natural sciences. In biology, specifications are independently given functional patterns that describe the goal-directed behavior of biological systems. A bacterial flagellum, for instance, is an outboard rotary motor on the backs of certain bacteria for propelling them through their watery environments. This functional description is epistemically objective but on any naturalistic construal of science must be regarded as ontologically subjective (if nature, as naturalism requires, is a closed nexus of undirected natural causes, then nature knows nothing about such functional descriptions). And yet biology as a science would be impossible without such functional descriptions. Functional language is indispensable to biology, and specifications are one way to clarify and make precise that language.
Any problem justifying specified complexity's legitimacy within science therefore resides elsewhere. Indeed, the problem resides with complexity. Although complexity becomes ontologically objective once a specification is in place, our assessment of complexity is just that -- our assessment. And the problem with assessments is that they can be wrong. Specifications are under our control. We formulate specifications on the basis of background knowledge. The complexity denoted by specified complexity, on the other hand, resides in nature. This form of complexity is a measure of probability, and these probabilities depend on the way nature is constituted. There is an objective fact of the matter what these probabilities are. But our grasp of these probabilities can be less than adequate. The problem, then, with specified complexity legitimately entering science is bridging complexity as it exists in nature with our assessments of that complexity. Alternatively, the problem is not with specified complexity being a valid property for science but with our ability to justify any particular attribution of specified complexity to particular objects or events in nature.
To illustrate what's at stake, consider an analogy from mathematics. There exist numbers whose decimal expansions are such that every single digit between 0 and 9 has relative frequency exactly 10 percent as the decimal expansion becomes arbitrarily large (or, as mathematicians would say, "in the limit" each single digit has exactly a 10 percent occurrence). The simplest such number is perhaps .01234567890123456789... where "0123456789" just keeps repeating over and over again. Let's call such numbers regular (mathematicians typically prefer a stronger notion of regularity called normality, which characterizes the limiting behavior of all finite strings of digits and not merely that of single digits; for the purposes of this example, however, regularity suffices). The property X is regular therefore applies to this number. Regularity is clearly a legitimate mathematical property -- it is perfectly well-defined and numbers either are regular or fail to be regular.
But suppose next we want to determine whether the number pi is regular (pi equals the ratio of the circumference of a circle to its diameter). Pi has a nonrepeating decimal expansion. Over the years mathematicians and computer scientists have teamed up to compute as many digits of pi as mathematical methods and computer technology permit. The current record stands at 206,158,430,000 decimal digits of pi and is due to the Japanese researchers Yasumasa Kanada and Daisuke Takahashi (the currently standard 40 gigabyte harddrive is too small to store this many decimal digits). Each of the single digits between 0 and 9 has relative frequency roughly 10 percent among these 200 billion decimal digits of pi. Is pi therefore regular?
Just as there is a physical fact of the matter whether an object or event in nature exhibits specified complexity, so there is a mathematical fact of the matter whether pi is regular. Pi either is regular or fails to be regular. Nonetheless, the determination whether pi is regular is another matter. With the number .01234567890123456789..., its regularity is evident by inspection. But the decimal digits of pi are nonrepeating, and to date there is no theoretical justification of its regularity. The closest thing to a justification is to point out that for the standard probability measure on the unit interval (i.e., Lebesgue measure), all numbers except for a set of probability zero are regular. The presumption, then, is that pi is likely to be regular. The problem here, however, is that the numbers we deal with in practice are rational numbers and most of these are not regular. Thus most of the numbers we deal with in practice belong to that set of probability zero. What's more, a simple set theoretic argument shows that among irrational numbers like pi, there are as many nonregular ones as regular ones (both subsets have cardinality of the continuum). There is thus no reason to think that pi was sampled according to Lebesgue probability measure and therefore likely to fall among the regular irrational numbers (the nonregular irrational numbers having probability zero with respect to Lebesgue measure). As a consequence, we have no basis in mathematical experience or theory for being confident that pi is regular.
Even the discovery that the single digits of pi have approximately the right relative frequencies for pi's first 200 billion decimal digits provides no basis for confidence that pi is regular. However regular the decimal expansion of pi looks in some initial segment, it could go haywire thereafter, possibly even excluding certain single-digits entirely after a certain point. On the other hand, however nonregular the decimal expansion of pi looks in some initial segment, the relative frequencies of the single digits between 0 and 9 could eventually settle down into the required 10 percent and pi itself could be regular (any initial segment thereby getting swamped by the infinite decimal expansion that lies beyond it). Thus, to be confident that pi is regular, mathematicians need a strict mathematical proof showing that each single digit between 0 and 9 has a limiting relative frequency of exactly 10 percent.
Now critics of intelligent design demand this same high level of justification (i.e., mathematical proof) before they accept specified complexity as a legitimate tool for science. Yet a requirement for strict proof, though legitimate in mathematics, is entirely wrong-headed in the natural sciences. The natural sciences make empirically based claims, and such claims are always falsifiable. Errors in measurement, incomplete knowledge, and the problem of induction cast a shadow over all scientific claims. To be sure, the shadow of falsifiability doesn't incapacitate science. But it does make the claims of science (unlike those of mathematics) tentative, and it also means that we need to pay special attention to how scientific claims are justified. The key question for this discussion, therefore, is how to justify ascribing specified complexity to natural structures.
To see what's at stake, consider further the analogy between the regularity of numbers and the specified complexity of natural structures. We need to be clear where that analogy holds and where it breaks down. The analogy holds insofar as both specified complexity and regularity make definite claims about some fact of the matter. In the case of regularity, it is a mathematical fact of the matter -- the decimal expansions of numbers either exemplify or fail to exemplify regularity. In the case of specified complexity, it is a physical fact of the matter -- a biological system, for instance, either exemplifies or fails to exemplify specified complexity. This last point is worth stressing. Attributing specified complexity is never a meaningless assertion. On the assumption that no design or teleology was involved in the production of some event, that event has a certain probability and therefore an associated measure of complexity. Whether that level of complexity is high enough to qualify the event as exemplifying specified complexity depends on the physical conditions surrounding the event. In any case, there is a definite fact of the matter whether specified complexity obtains.
Any problem with ascribing specified complexity to that event therefore resides not in its coherence as a meaningful concept -- specified complexity is well-defined. If there is a problem, it resides in what philosophers call its assertibility. Assertibility refers to our justification for asserting the claims we make. A claim is assertible if we are justified asserting it. With the regularity of pi, it is possible that pi is regular. Thus in asserting that pi is regular, we might be making a true statement. But without a mathematical proof of pi's regularity, we have no justification for asserting that pi is regular. The regularity of pi is, at least for now, unassertible. But what about the specified complexity of various biological systems? Are there any biological systems whose specified complexity is assertible?
Critics of intelligent design argue that no attribution of specified complexity to any natural system can ever be assertible. The argument runs as follows. It starts by noting that if some natural system exemplifies specified complexity, then that system must be vastly improbable with respect to all purely natural mechanisms that could be operating to produce it. But that means calculating a probability for each such mechanism. This, so the argument runs, is an impossible task. At best science could show that a given natural system is vastly improbable with respect to known mechanisms operating in known ways and for which the probability can be estimated. But that omits (1) known mechanisms operating in known ways for which the probability cannot be estimated, (2) known mechanisms operating in unknown ways, and (3) unknown mechanisms.
Thus, even if it is true that some natural system exemplifies specified complexity, we could never legitimately assert its specified complexity, much less know it. Accordingly, to assert the specified complexity of any natural system constitutes an argument from ignorance. This line of reasoning against specified complexity is much like the standard agnostic line against theism -- we can't prove that atheism (cf. the total absence of specified complexity from nature) holds, but we can show that theism (cf. the specified complexity of certain natural systems) cannot be justified and is therefore unassertible. This is how skeptics argue that there is no (and indeed can be no) evidence for God or design.
A little reflection, however, makes clear that this attempt by skeptics to undo specified complexity cannot be justified on the basis of scientific practice. Indeed, the skeptic imposes requirements so stringent that they are absent from every other aspect of science. If standards of scientific justification are set too high, no interesting scientific work will ever get done. Science therefore balances its standards of justification with the requirement for self-correction in light of further evidence. The possibility of self-correction in light of further evidence is absent in mathematics and accounts for mathematics' need for the highest level of justification, namely, strict logico-deductive proof. But science does not work that way. Science must work with available evidence and on that basis (and that basis alone) formulate the best explanation of the phenomenon in question.
Take, for instance, the bacterial flagellum. Despite the thousands of research articles on it, no mechanistic account of its origin exists. Consequently, there is no evidence against its being complex and specified. It is therefore a live possibility that it is complex and specified. But is it fair to assert that it is complex and specified, in other words, to assert that it exhibits specified complexity? The bacterial flagellum is irreducibly complex, meaning that all its components are indispensable for its function as a motility structure. What's more, it is minimally complex, meaning that any structure performing the bacterial flagellum's function as a bidirectional outboard rotary motor cannot make do without certain basic components.
Consequently, no direct Darwinian pathway exists that incrementally adds these basic components and therewith evolves a bacterial flagellum. Rather, an indirect Darwinian pathway is required, in which precursor systems performing different functions evolve by changing functions and components over time (Darwinists refer to this as coevolution and co-optation). Plausible as this sounds (to the Darwinist), there is no evidence for it. What's more, evidence from engineering strongly suggests that tightly integrated systems like the bacterial flagellum are not formed by trial and error tinkering in which form and function coevolve. Rather, such systems are formed by a unifying conception that combines disparate components into a functional whole -- in other words, by design.
Does the bacterial flagellum exhibit specified complexity? Is such a claim assertible? Certainly the bacterial flagellum is specified. One way to see this is to note that humans developed outboard rotary motors well before they figured out that the flagellum was such a machine. This is not to say that for the biological function of a system to constitute a specification, humans must have independently invented a system that performs the same function. Nevertheless, independent invention makes all the more clear that the system satisfies independent functional requirements and therefore is specified. At any rate, no biologist I know questions whether the functional systems that arise in biology are specified. At issue always is whether the Darwinian mechanism, by employing natural selection, can overcome the vast improbabilities that at first blush seem to arise with such systems, thereby breaking a vast improbability into a sequence of more manageable probabilities.
To illustrate what's at stake in breaking vast improbabilities into more manageable probabilities, suppose a hundred pennies are tossed. What is the probability of getting all one hundred pennies to exhibit heads? The probability depends on the chance process by which the pennies are tossed. If, for instance, the chance process operates by tossing all the pennies simultaneously and does not stop until all the pennies simultaneously exhibit heads, it will require on average about a thousand billion billion billion such simultaneous tosses for all the pennies to exhibit heads. If, on the other hand, the chance process tosses only those pennies that have not yet exhibited heads, then after about eight tosses, on average, all the pennies will exhibit heads. Darwinists tacitly assume that all instances of biological complexity are like the second case, in which a seemingly vast improbability can be broken into a sequence of reasonably probable events by gradually improving on an existing function (in the case of our pennies, improved function would correspond to exhibiting more heads).
Irreducible and minimal complexity challenge the Darwinian assumption that vast improbabilities can always be broken into manageable probabilities. What evidence there is suggests that such instances of biological complexity must be attained simultaneously (as when the pennies are tossed simultaneously) and that gradual Darwinian improvement offers no help in overcoming their improbability. Thus, when we analyze structures like the bacterial flagellum probabilistically on the basis of known material mechanisms operating in known ways, we find that they are highly improbable and therefore complex in the sense required by specified complexity.
Is it therefore fair to assert that the bacterial flagellum exhibits specified complexity? Design theorists say yes. Evolutionary biologists say no. As far as the evolutionary biologists are concerned, design theorists have failed to take into account indirect Darwinian pathways by which the bacterial flagellum might have evolved through a series of intermediate systems that changed function and structure over time in ways that we do not yet understand. But is it that we do not yet understand the indirect Darwinian evolution of the bacterial flagellum or that it never happened that way in the first place? At this point there is simply no evidence for such a indirect Darwinian evolutionary pathways to account for biological systems that display irreducible and minimal complexity.
Is this, then, where the debate ends, with evolutionary biologists chiding design theorists for not working hard enough to discover those (unknown) indirect Darwinian pathways that lead to the emergence of irreducibly and minimally complex biological structures like the bacterial flagellum? Alternatively, does it end with design theorists chiding evolutionary biologists for deluding themselves that such indirect Darwinian pathways exist when all the available evidence suggests that they do not. Although this may seem like an impasse, it really isn't. Like compulsive gamblers who are constantly hoping that some big score will cancel their debts, evolutionary biologists live on promissory notes that show no sign of being redeemable. Science must form its conclusions on the basis of available evidence, not on the possibility of future evidence. If evolutionary biologists can discover or construct detailed, testable, indirect Darwinian pathways that account for the emergence of irreducibly and minimally complex biological systems like the bacterial flagellum, then more power to them -- intelligent design will quickly pass into oblivion. But until that happens, evolutionary biologists who claim that natural selection accounts for the emergence of the bacterial flagellum are worthy of no more credence than compulsive gamblers who are forever promising to settle their accounts.
There is further reason to be skeptical of evolutionary biology and side with intelligent design. In the case of the bacterial flagellum, what keeps evolutionary biology afloat is the possibility of indirect Darwinian pathways that might account for it. Practically speaking, this means that even though no slight modification of a bacterial flagellum can continue to serve as a motility structure, a slight modification could serve some other function. But there is now mounting evidence of biological systems for which any slight modification does not merely destroy the system's existing function but also destroys the possibility of any function of the system whatsoever (cf. Xxx Yyy's work on individual enzymes). For such systems, neither direct nor indirect Darwinian pathways could account for them. In this case we would be dealing with an in-principle argument showing not merely that no known material mechanism is capable of accounting for the system but also that any unknown material mechanism is incapable of accounting for it as well. The argument here turns on an argument from contingency and degrees of freedom outlined in the previous chapter.
Is the claim that the bacterial flagellum exhibits specified complexity assertible? You bet! Science works on the basis of available evidence, not on the promise or possibility of future evidence. Our best evidence points to the specified complexity (and therefore design) of the bacterial flagellum. It is therefore incumbent on the scientific community to admit, at least provisionally, that the bacterial flagellum could be the product of design. Might there be biological examples for which the claim that they exhibit specified complexity is even more assertible? Yes there might. Assertibility comes in degrees, corresponding to the strength of evidence that justifies a claim. For the bacterial flagellum it is logically impossible to rule out the infinity of possible indirect Darwinian pathways that might give rise to it (though any such proposed route to the bacterial flagellum is at this point a mere conceptual possibility). Yet for other systems, like certain enzymes, there can be strong grounds for ruling out such indirect Darwinian pathways as well.
The evidence for intelligent design in biology is therefore destined to grow ever stronger. There's only one way evolutionary biology could defeat intelligent design, and that is by in fact solving the problem that it claimed all along to have solved but in fact never did, namely, to account for the emergence of multi-part tightly integrated complex biological systems (many of which display irreducible and minimal complexity) apart from teleology or design. To claim that the Darwinian mechanism solves this problem is false. The Darwinian mechanism is not itself a solution but rather describes a reference class of candidate solutions that purport to solve this problem. But none of the candidates examined to date indicates the slightest capacity to account for the emergence of multi-part tightly integrated complex biological systems. That's why molecular biologist James Shapiro, who is not a design theorist, writes, "There are no detailed Darwinian accounts for the evolution of any fundamental biochemical or cellular system, only a variety of wishful speculations." (Quoted from his 1996 book review of Darwin's Black Box that appeared in The National Review.)
In summary, specified complexity is a well-defined property that applies meaningfully to events and objects in nature. Specified complexity is an objective property -- specifications are epistemically objective and complexity is ontologically objective. Any concern over specified complexity's legitimacy within science rests not with its coherence or objectivity, but with its assertibility, namely, with whether and the degree to which ascribing specified complexity to some natural object or event is justified. Any blanket attempt to render specified complexity unassertible gives an unfair advantage to naturalism, ensuring that design cannot be discovered even if it is present in nature. What's more, science can proceed only on available evidence, not on the promise or possibility of future evidence. As a consequence, ascriptions of specified complexity to natural objects and events, and to biological systems in particular, can be assertible. And indeed, there are actual biological systems for which ascribing specified complexity -- and therefore design -- is eminently assertible.
IP: Logged
|
|
Frances
Member
Member # 169
|
posted 30. August 2002 02:51
Frances post has been deleted. He seems to have been waiting for an opportunity to unleash his general criticisms of Dembski and intelligent design. Unfortunately, Dembski's post here is on specified complexity as a property and it is that which needs to be addressed. Frances post was completely off topic.
===
I would not say that my comments were completely off topic since I addressed several claims made by Dembski.
I will abide by the moderator's decision and focus even more closely on the topic at hand and on Dembski's claims:
The claim that specified complexity is not subjective seems to be undermined by the statement that the flagellum exhibits specification since it mimics an outboard motor and that it is minimally complex as such. But the first claim is a subjective claim and the second claim presumes that the function "outboard motor" is the only function possible for the intermediates of the flagellum.
What I find interesting is that Dembski asserts that science should deal with the available evidence and yet the ID inference is made on the absence of available evidence. Perhaps the comparison of compulsive gambling would equally well or better apply to the IDer who hopes that no such future pathways to an IC system exist, despite the logical possibility of their existence. It was just recently that creationists claimed that no pathways existed for the middle ear to arise in mammals that evolved from reptiles. And yet science uncovered exactly those intermediates to show how pre-existing structures in the jaw bones were co-opted to function as impedance matchers in the ear.
=====Addition=====
What I have found even more intruiging is that Dembski's argument about IC does not rule out indirect pathways anymore. This might open up the use of the term complexity and specification to allow us to research how both intelligent design and natural design use similar and dissimilar approaches to generate CSI.
[ 31 August 2002, 12:39: Message edited by: Frances ]
IP: Logged
|
|
warren_bergerson
Member
Member # 262
|
posted 30. August 2002 13:29
While it is interesting to review the philosophical aspects of the specified complexity property, I personally would be more interested in the practical-scientific aspects of the property. I would be interested in seeing answers to questions such as:
1. On what objective basis can we or could be reach general/technical agreement on measures of complexity?
2. On what objective basis can we or could we reach general/technical agreement on the characteristic that separates specified from non-specified complexity?
3. What scientific issue or issues is specified complexity designed to answer?
4. Can specified complexity be defined in terms of or reduced in part or whole to existing quantifiable terms and concepts?
5. Can the presence or absence of the specified complexity property be determined for any real world phenomena?
6. Are there real world objects on the ‘boarder’ between specified and non-specifed complexity?
7. Would or could the definition of the specified complexity property change if the prevalent theory of evolution were to change?
While it is both admirable and interesting to address the philosophy issues surrounding specified complexity, IMO, it would be more useful to answer the questions of the technician trying to build the specified complexity thermometer. Let me also say that, IMO, design theory or design science has advanced to point where the technicians questions could be answered.
IP: Logged
|
|
michaelgoodrich
Member
Member # 393
|
posted 01. September 2002 11:17
I invite Frances to consider that the lack of evidence for something is valid evidence of its non-existence, at least provisionally, the dangers of the Argument from Ignorance fallacy notwithstanding.
This situation is more acute when there exists a reasonable alternative.
I also invite him to consider that, while intelligent design is indisputably true as a feature of the Universe, curiously it is mainly controversial only when applied to biological systems.
Conversly the concept of 'natural design' is conjectural and no one knows if there is actually any such thing.
[ 01 September 2002, 11:38: Message edited by: michaelgoodrich ]
IP: Logged
|
|
Frances
Member
Member # 169
|
posted 01. September 2002 13:49
Michael considers lack of evidence, evidence of its non-existance. But ID goes beyond that by claiming that it thus is evidence of intelligent design.
Is ID a reasonable alternative? I would argue that in absence of mechanism, historical pathway etc it does not seem to be better than admitting that we don't know.
Intelligent design is controversial when it applies to biological systems because there is no independent evidence of an intelligent designer and there are countless alternatives that do not require such intelligent designer.
Natural design may be conjectural but it is based on real mechanmisms. Seems that natural design in biology has an enormous advantage over its 'alternative'. For instance Dembski in his chapter above allows for indirect pathways being able to generate IC systems, thus allowing for natural ways to generate CSI. This seems to be a major departure from the CSI claims in NFL. In fact as Howard van Till in E. COLI AT THE NO FREE LUNCHROOM Bacterial Flagella and Dembski’s Case for Intelligent Design seems to identify some problems with the bacterial flagellum as evidence of ID.
van Till takes both the compexity and specificity claim for the flagellum and applies Dembski's criteria.
As for instance Tom Schneider has shown in Evolution of biological information Darwinian mechanisms such as mutation and selection can generate biological information. Surely this should be considered evidence of a natural design?
Other examples of the use of natural design can be found in genetic programming
quote:
This presentation will give an overview of evolutionary design in nature, and how specific natural design principles such as selection, mutation, and emergence can be better understood through mathematical and computational models. Demonstrations of how to breed computer programs and how to utilize emergent computation through swarms will illustrate the usefulness and potential of these algorithmic approaches gleaned from nature.
[ 01 September 2002, 14:50: Message edited by: Frances ]
IP: Logged
|
|
Kirk Durston
Member
Member # 174
|
posted 02. September 2002 12:30
I want to respond to some comments made by Warren and Francis. Warren's questions first. Ultimately, evolution must take place on the level of proteins and the genetic regulatory system that governs their expression. The specified complexity (which I will refer to as simply 'specificity') of both proteins and genetic regulatory systems can be mathematically defined, it seems to me. If one takes the difference in Shannon entropy between a given sequence of amino acids, with no regard as to whether it is functional or not, and the Shannon entropy of a functional protein with the same number of residues, the resulting equation contains the ratio Nf/N where Nf = number of functional sequences and N = total number of sequences both functional and non-functional. The ratio Nf/N is, as I understand it, what Dembski is referring to when he speaks of specified complexity. Simply put, it represents the probability of obtaining a functional sequence for any given protein. It is also an objective indicator of how tightly the physical system must be constrained (specified) in order to obtain a functional protein.
Specificity defined as such seems to be entirely objective, as the ratio is determined by physiochemical factors completely independent of the existence of any minds, human or otherwise. It is also useful to the molecular biologists who is interested in constructing functional enzymes in the lab as it gives her an idea or whether or not she should rely on random sequence libraries. Incidentally, a paper by Taylor et al., rules out random sequence libraries on the basis of the highly constrained specificity of active enzymes.
This definition also seems to be independent of theoretical approaches to the origin and complexity of life.
Now for Francis' criticism of the gambler analogy in Dembski's discussion of assertability. Dembski mentions in his paper the work of Xxx Yyy on individual enzymes. I am not familiar with this specific work, but I am aware of several papers now published that indicate that the evolution of a functional protein can only proceed a limited amount before the fold of that protein becomes unstable and, hence, non-functional in the cell. What these papers are showing is that information in functional proteins is quantized, isolated to miniscule areas of sequence space. Thus it is unlikely that one protein can evolve into a completely novel fold via a series of functional, folded intermediates (Blanco et al.). To summarize, the empirical evidence is increasingly indicating two things. First, that specificity for functional proteins is constrained to such a degree that obtaining functional proteins via a random walk is not feasible within the lifetime of the universe. Second, functional sequence space appears to take the form of isolated islands within an ocean of non-folding sequence space. Thus, there are no pathways between the islands that can be guided by natural selection since natural selection cannot work with non-functional, non-folding proteins since they are not expressed in the phenotype. My point (and I think Dembski's point) is this. If there is no evidence novel folds can evolve through natural processes, and, in fact, increasing evidence that they cannot, and it is an empirical fact that intelligent agents can produce at least moderately active enzymes (Taylor et al.), which is the more rational position to take? Natural processes or ID? Thus Dembski's gambler analogy is entirely valid.
Please note that the argument for ID is not an argument from silence. It is an empirical fact that ID can produce specified complexity (e.g., Dembski's outboard motor). Thus ID is a valid option. Logically, it boils down to a valid dysjunction of the form:
Either natural processes or ID produced organic life.
This is a valid dyjunction provided there isn't a third option, which there isn't. Thus, if the empirical evidence increasingly indicates that natural processes cannot do the job then, necessarily, ID is required. To defeat this argument, the naturalist must show that ID cannot do it or that it is more rational to hold that natural processes did it. The naturalist can do neither, and the situation is worsening, not improving for the naturalist. Of course, the argument for ID is not limited to the above, there is positive evidence for ID, but that is outside the topic at hand.
IP: Logged
|
|
warren_bergerson
Member
Member # 262
|
posted 02. September 2002 13:32
Kirk,
I believe your comment provide a useful starting point in discussing/defining the issue of specified complexity. However, I believe, the issue is a bit more complicated than you have described. As you define it, N/Nf is an objective measure of complexity. This would also be a measure of ‘evolutionary design complexity’ if it was known or could be demonstrated that Nf was generated as a result of a random trial and error search of the space N. The creation or evolution of Nf could involve much less design complexity if Nf resulted from a ‘directed design process’ that searched a much smaller set of possible sequences. [It is also conceivable that the ‘design complexity’ is greater than N/Nf if the evolutionary/creation complex is more complex.]
There is, IMO, no reason to doubt that it is possible to develop objective measures of complexity. The measure will, however, be ‘relative to the question being analyzed’ rather than an absolute value. If specified complexity is intended to evaluate the ‘probability of design resulting from a random RM&NS process’, that is one thing. If specified complexity is also intended to evaluate the ‘probability of design resulting from a directed design process’ that is something very different.
As I stated before, I think the technical issues associated with specified complexity can be answered. Answering the technical issues, however, requires addressing both how the measure is produced and how the measure is to be used.
IP: Logged
|
|
charlie d.
Member
Member # 159
|
posted 02. September 2002 14:03
quote:
What these papers are showing is that information in functional proteins is quantized, isolated to miniscule areas of sequence space. Thus it is unlikely that one protein can evolve into a completely novel fold via a series of functional, folded intermediates (Blanco et al.). To summarize, the empirical evidence is increasingly indicating two things. First, that specificity for functional proteins is constrained to such a degree that obtaining functional proteins via a random walk is not feasible within the lifetime of the universe. Second, functional sequence space appears to take the form of isolated islands within an ocean of non-folding sequence space. Thus, there are no pathways between the islands that can be guided by natural selection since natural selection cannot work with non-functional, non-folding proteins since they are not expressed in the phenotype.
That the origin of new protein folds cannot be a common evolutionary occurrence is well known, clearly highlighted by the fact that there are at most a few thousands recognizable folds among all the known proteins, and in fact the vast majority of proteins (>80%, if I remember correctly) use only a few hundreds of them.
However, it is not true that we have no idea how one fold can evolve from another, in fact there are a number of cases in which homologous proteins display different structural folds, and such transitions can be ascribed to well-defined sequence alterations (that is, mutations). Here is an excellent review on the issue, which highlights a number of mechanisms that can cause fold transitions, such as insertion/deletion, strand invasion etc.
At least from the cases presented in the review, transitions seem to be due to admittedly rare, but run-of-the-mill molecular mechanisms. That protein evolution does not necessarily occur by gradual, aminoacid-by-aminoacid, transitions, and in fact is at times almost - gasp! - saltational in nature (domain swaps, protein fusions etc are commonplace) is also well known, so the "random walk with non-functional intermediates" issue is really a red herring.
The rarity of mutation events capable of causing fold transition would explain why so few folds exist in nature. Of course, that such transitions can occur does not say anything about how they occurred, whether by design or RM/NS. However, at the current state of knowledge they certainly are not incompatible with naturalistic, darwinian evolutionary models.
[ 02 September 2002, 14:15: Message edited by: charlie d. ]
IP: Logged
|
|
Frances
Member
Member # 169
|
posted 02. September 2002 14:07
There is no doubt that ID can produce specified complexity although I find the outboard motor example totally irrelevant since it does not apply to biology. The question really is can natural design result in specified complexity? So far the evidence seems to suggests yes and certainly there is no apriori reason to reject this possibility.
As far as protein evolution is concerned, at most it has been shown that folded intermediates may not have a selective advantage thus Darwinian evolution of proteins may or may not be possible.
Does ID provide a pathway, a mechanism? The claim is that there are two possibilities, natural design or intelligent design. I would add a third one, we don't know yet.
Examples of hypotheses of protein evolution include
A hierarchical approach to protein molecular evolution
Protein topology and stability define the space of allowed sequences
Novel folded protein domains generated by combinatorial shuffling of polypeptide segments
Roles of mutation and recombination in the evolution of protein thermodynamics
To show some examples of how evolutionary mechanisms can generate CSI see for instance Radio emerges from the electronic soup
quote:
A self-organising electronic circuit has stunned engineers by turning itself into a radio receiver.
What should have been an oscillator became a radio This accidental reinvention of the radio followed an experiment to see if an automated design process, that uses an evolutionary computer program, could be used to "breed" an electronic circuit called an oscillator. An oscillator produces a repetitive electronic signal, usually in the form of a sine wave.
...
These "fittest" candidates were then mated by mixing their genes together, or mutated by making random changes to them.
Evolvable hardware is a fascinating application of natural design.
[ 02 September 2002, 14:56: Message edited by: Frances ]
IP: Logged
|
|
Kirk Durston
Member
Member # 174
|
posted 02. September 2002 16:44
Re. Warren's comments: it should be noted that the ratio Nf/N is completely independent of pathways, evolutionary design or directed design by intelligence. Rather than Nf/N being relative to the question being analyzed, it is relevant only to physiochemical function, in the case of proteins. Since it is laws of physics that determine protein folding, subjective human thinking and theories of how a functional protein came about are irrelevant to specificity. The question we wrestle with is what process can produce the specified complexity of a given protein.
Re. Charlie's comments: I need to point out that I did not assert that we have no idea how one protein could evolve into another, clearly a series of mutations could do it given enough time and opportunities. It is easy to look at two proteins which we theorize to be homologous and construct a mutational pathway to achieve the orthologous or paralogous protein. The problem arises when we calculate the probability of achieving the novel protein and then realize that it is too small. A recently published paper in Phys. Rev. Lett. shows that the most number of operations that could have occurred in the universe to date is 10^120. What that indicates is that the finest specificity we could reasonably expect nature to achieve is about 10^-120. The average 300 residue protein has a specificity that is considerably finer than that. My point is that nature can produce specified complexity but the upper limit for such is around 10^-120.
Regarding domain swaps, protein fusion, etc., we have a problem. Specificity happens to be the primary component of Shannon's equation for the information of functional systems. The upper limit for the amount of functional information natural processes can produce on a macroscopic scale (scales that are large enough such that quantum effects become negligible) is about 70 bits. Thus if a novel protein can be produced by processes that require less than 70 bits, then it is possible that natural processes could produce it. A case in point are antigen receptor proteins. Unfortunately, there is a problem with this approach. One of the principle contributors to the discussion after Claude Shannon’s paper was Leon Brillouin. Brillouin pointed out that the total information required to produce a functional sequence or configuration, is composed of two parts (Brillouin, 1951). Part A is “the part corresponding to the technical organization of the system of transmission”. Part B is the information contained in the message itself. Applied to proteins, Part A includes, among other things, the pre-existing information required to produce the necessary requirements for the novel protein to be achieved. Part B is the novel functional information actually carried by the protein. In the case of antigen receptor proteins, Part A information includes the functional information required to produce the pre-requisite DNA, the RAG protein complex for cutting, and the protein complex for recombination. Part A information would be very high, well above 70 bits.
For domain swaps, protein fusion, etc., the problem is in obtaining the existing Part A information. Once that problem has been overcome, then the Part B information is relatively small and achievable. In every case I have investigated, and I continue to do so, the part A information (in other words, the specificity of the raw materials) is far to high to be achieved through natural processes. So domain swaps, protein fusion, etc., represent the rearrangement of existing functional sequences, which require little, if any additional functional information. I would be happy to respond to a specific case (time constraints prevent me from addressing all the papers referred to by Charlie and Frances). If you are aware of the creation of a novel protein that does not require a prohibitively large amount of part A information, post it here and I'll take a look at it. I certainly have not found any yet.
Regarding random walks as a red herring, it is Blanco's conclusion, not to mention it is patently obvious, that the evolutionary trajectories toward completely novel folds will reduce to a random walk if they must occur by mutation and if the intervening sequence space is non-folding.
Re. Charlie's statement, " The rarity of mutation events capable of causing fold transition would explain why so few folds exist in nature." I would dispute that the number of folds in nature is a 'few'. It is postulated that the bare-bones minimum number of proteins necessary for life is about 150, although the general consensus is that 300 is more likely. Higher life forms appear to contain thousands. The specificity of the simplest bare-bones genome is approximately 10^-12,600. Given the upper limit for natural specified complexity is around 10^-120, naturalists have a problem. However, even if organic life required only one 300 residue protein, we would still have a very large problem in explaining how natural processes could produce such.
Re. Francis' assertion that the 'evidence seems to suggest yes' regarding whether or not natural processes can produce specified complexity: As I've indicated above, natural processes can produce specified complexity but there is an upper limit. For all events, including quantum events, that limit is 10^-120. For non-quantum events, the limit is likely to be closer to 10^-21. Unfortunately for the naturalist, even a moderately active enzyme requires a specificity finer than this, as Taylor et al. have shown.
Regarding Francis' third option 'we don't know yet', if we don't even have a concept of what might fit into this third category, then it isn't even an option. Keep in mind that 'natural design' includes the set of all possible natural processes, known and unknown. The third option cannot include any natural processes. The second option includes all ID, human and otherwise. The third option cannot include that. So, can Francis even conceive of what that third option might be? If not, then we are left with a valid dysjunction. It is not what we do not know that is the problem, in the origin of functional proteins, but it is what we do know that is the problem, specifically, their specificity and the miniscule fraction of sequence space that produces stable folds.
I should point out that it is beginning to look as if explaining proteins will be relatively simple in comparison with understanding and explaining the origin of the entire regulatory system, which appears to function as an operating system for which proteins are merely standardized tools. My point is that rather than the gap narrowing as we do further research, it is expanding at an alarming rate.
IP: Logged
|
|
Frances
Member
Member # 169
|
posted 02. September 2002 17:08
I am somewhat confused by the term 'specificity'. What relevance does specificity have to the discussion at hand? An example which may help clarify my confusion is 500 coin tosses. The 'specifity' of the outcome seems to be about ~10^-150. Yet I would not argue that this presents a problem to coin tosses being valid chance hypotheses.
If the argument is that proteins perform a specific function then we need to establish more than just a random chance 'creation' of such. Perhaps we should define the term specificity for this discussion.
Also if we accept the definition of natural design to include known and unknown processes, the my argument remains as before, how can we reject natural design in presence of our ignorance? Would not even ID as it present stands wrt biology fall in the category of 'we don't know'?
[ 02 September 2002, 21:17: Message edited by: Frances ]
IP: Logged
|
|
charlie d.
Member
Member # 159
|
posted 02. September 2002 18:03
quote:
Regarding domain swaps, protein fusion, etc., we have a problem. Specificity happens to be the primary component of Shannon's equation for the information of functional systems. The upper limit for the amount of functional information natural processes can produce on a macroscopic scale (scales that are large enough such that quantum effects become negligible) is about 70 bits. Thus if a novel protein can be produced by processes that require less than 70 bits, then it is possible that natural processes could produce it. A case in point are antigen receptor proteins. Unfortunately, there is a problem with this approach. One of the principle contributors to the discussion after Claude Shannon’s paper was Leon Brillouin. Brillouin pointed out that the total information required to produce a functional sequence or configuration, is composed of two parts (Brillouin, 1951). Part A is “the part corresponding to the technical organization of the system of transmission”. Part B is the information contained in the message itself. Applied to proteins, Part A includes, among other things, the pre-existing information required to produce the necessary requirements for the novel protein to be achieved. Part B is the novel functional information actually carried by the protein. In the case of antigen receptor proteins, Part A information includes the functional information required to produce the pre-requisite DNA, the RAG protein complex for cutting, and the protein complex for recombination. Part A information would be very high, well above 70 bits.
For domain swaps, protein fusion, etc., the problem is in obtaining the existing Part A information. Once that problem has been overcome, then the Part B information is relatively small and achievable. In every case I have investigated, and I continue to do so, the part A information (in other words, the specificity of the raw materials) is far to high to be achieved through natural processes. So domain swaps, protein fusion, etc., represent the rearrangement of existing functional sequences, which require little, if any additional functional information.
Honestly, you lost me at the 70 bits (I can probably only process 69 at a time ![[Wink]](wink.gif) ). I do however understand antigen receptor genes, so if you care to explicate that issue in biological terms, I'll be happy to discuss it. As for the "domain swap" thing, I was just mentioning it as an example of "saltational" evolution of proteins - I do agree that these mechanisms do not generate new "information" in absolute, IT terms, although of course they generate novel biological functions, which is all evolution really cares about.
quote:
Regarding random walks as a red herring, it is Blanco's conclusion, not to mention it is patently obvious, that the evolutionary trajectories toward completely novel folds will reduce to a random walk if they must occur by mutation and if the intervening sequence space is non-folding.
Exactly. What is a red herring is to say that evolution of proteins has to follow necessarily such a random walk. When it can, it does (and most of the evolution of proteins within structurally related families, by duplication/divergence, probably occurs this way); however, because saltational mechanisms are able to a) generate new functional combinations from old ones (as in domain swaps), and b) generate new functionally coherent structural folds (as described in the review I quote above), there is no reason to see the limitations of random walks as a significant problem for protein fold evolution. In other words, all that Blanco's results show is that gradual evolution of new protein domains by random walk is unlikely, not the evolution of new protein domains is unlikely.
quote:
Re. Charlie's statement, " The rarity of mutation events capable of causing fold transition would explain why so few folds exist in nature." I would dispute that the number of folds in nature is a 'few'. It is postulated that the bare-bones minimum number of proteins necessary for life is about 150, although the general consensus is that 300 is more likely. Higher life forms appear to contain thousands. The specificity of the simplest bare-bones genome is approximately 10^-12,600. Given the upper limit for natural specified complexity is around 10^-120, naturalists have a problem. However, even if organic life required only one 300 residue protein, we would still have a very large problem in explaining how natural processes could produce such.
I think you are confusing terms here. "Fold", or "domain" indicate a structurally defined region in a protein, that is usually shared by several different proteins, and can be regognized on the basis of specific sequence motifs. For instance, the Ig fold, first identified in antibody molecules, is a globular domain of 110-120 aa residues, characterized by 2 cysteines forming a disulfide bond, and 2 anti-parallel beta sheets facing each other. Such a domain is present in hundreds of proteins, mostly working as receptors of sort, but otherwise fucntionally unrelated to antibodies. Similarly with other kinds of domain. Also, each protein can present multiple unrelated domains. If you look at all the proteins whose sequence is known and structure known or inferred, you'll find that most proteins only use a very limited array of domains (a few hundreds), while other domains are present in only a few proteins. Altogether, compared to the enormous variability of proteins and organisms, the total number of recognized domains (in the thousands) is rather puny.
In evolutionary terms, this suggests that while evolution within each domain family (by duplication/divergence, mostly) is rather common, the generation of entirely new protein domains is a considerably more rare event. Again, this is assuming that proteins are related by descent (if your version of ID assumes that each protein is created ex novo, I am sure you'd reach quite different conclusions). Nevertheless, from all we know, empirically, about how proteins can evolve, neither gradual evolution of one member of a structural family into another with different function, nor the occasional generation of entirely new structural motifs (folds/domains) seem to be insormountable barriers to naturalistic processes.
[ 02 September 2002, 18:04: Message edited by: charlie d. ]
IP: Logged
|
|
peroxisome
unregistered
|
posted 02. September 2002 18:38
[***Moderator Merge: First Post***]
quote:
the resulting equation contains the ratio Nf/N where Nf = number of functional sequences and N = total number of sequences both functional and non-functional
hmmm, I smell something odoriferous. Presumably you can only do this calculation if you know Nf; and who has made a systematic study of Nf for large proteins? Oh- it hasn't been done, because it cannot be done because of the sheer numbers involved. And there are excellent examples of multi-gene families - cytochromes P450, where there are >100 families of P450, all of which have less than 40% similarity and greater than 25%. This indicates the sequence area which can produce a functional protein is not trivial.
So this objective measure of protein complexity may not be so very objective. And I am surprised ?
[***Moderator Merge: Second Post***]
quote:
Incidentally, a paper by Taylor et al., rules out random sequence libraries on the basis of the highly constrained specificity of active enzymes.
I don't suppose you would like to give the reference, would you, just so I can see exactly what the paper says ?
Quite bizarre, really, that there are so many groups who have used random mutagenesis, and who are continually improving the methodology. I count 850 citations in the last few years, and here is an example:
Directed evolution of selective enzymes and hybrid catalysts, Reetz MT, TETRAHEDRON, 58 (32): 6595-6602 AUG 5 2002
Can't see Dr Reetz giving up in disgust. ![[Smile]](smile.gif)
[***Moderator Merge: Third Post***]
quote:
It is an empirical fact that ID can produce specified complexity (e.g., Dembski's outboard motor). Thus ID is a valid option. <snip>
Either natural processes or ID produced organic life.
can you please give me an example of someone who has designed organic life de novo ?
Presumably you will also accept that it is an empirical fact that ID has not produced some forms of sufficiently complex specified complexity ?
per
[ 02 September 2002, 20:31: Message edited by: Moderator ]
IP: Logged
|
|
Moderator
Administrator
Member # 1
|
posted 02. September 2002 20:27
peroxisome,
Please try to keep from making short posts. You can do us all a great service by merging all your thoughts at a particular time into one post.
Thanks.
IP: Logged
|
|
Art
Member
Member # 179
|
posted 02. September 2002 20:30
quote:
Oh- it hasn't been done, because it cannot be done because of the sheer numbers involved. And there are excellent examples of multi-gene families - cytochromes P450, where there are >100 families of P450, all of which have less than 40% similarity and greater than 25%. This indicates the sequence area which can produce a functional protein is not trivial.
So this objective measure of protein complexity may not be so very objective.
I'd add my agreement with the suggestion made by peroxisome, and note that P450's are excellent examples of the inherently low information content of proteins. This is something that has been corroborated by many, many experiments.
However, as I argue here, complexity, as defined in terms of probability, really has nothing to do with absolute values of informational content.
I'd also add that the repeated references to work such as that by Blanco et al. are a bit confusing, since no one (that I am aware of, at least) has ever proposed that all extant proteins arose by progressive, step-by-step means from a single hypothetical ancestor. Moreover, many empirical studies have been pointed out by others that argue against the claims made by some ID advocates; I'd add the paper I menton to Jed Macasko in this thread as a demonstration of the surprising multiplicity of pathways that are actually available for the evolution of proteins.
Given all of this, it's hard to see how to derive a design-based model from studies such as that by Blanco et al. It's even harder to see an anti-evolutionary significance, which is strongly implied in much of the discussion.
IP: Logged
|
|
|
Printer-friendly view of this topic