Rube Goldberg Machines and Evolution

Rube Goldberg Machines

Evolution

Sean D. Pitman M.D.

© June 2002

In chapter four of his controversial book, Darwin's Black Box, Michael Behe presents the argument that Rube Goldberg machines exist in living things and that such machines are "irreducibly complex." Behe presents the argument that the existence of such machines cannot be explained by mindless naturalistic mechanisms and are thus examples of deliberate design. After all, anyone who has watched cartoons as a child knows what a Rube Goldberg machine is and that this machine will not work if any one part is removed. So, how can something evolve in a stepwise way where each step is functionally beneficial if there is no function until all the parts are in place? As an example, consider the following scenario where Behe describes a popular cartoon about the loud-mouthed rooster Foghorn Leghorn.

"...Foghorn would be walking along, notice a dollar bill or some other bait on the ground, and pick it up. The dollar was tied by a string to a stick that was propped against a ball. When the dollar bill was moved, the attached string pulled down the stick, and the ball would start to roll away as Foghorn stared slack-jawed at the developing action. The ball then would fall of a cliff onto the raised end of a seesaw, smacking it down and sending a rock with an attached piece of sandpaper hurtling into the air. On its upward journey the sandpaper would strike a match sticking out of the cliff, which lit the fuse to a cannon. The cannon would fire; on its downward track the cannonball would hit the rim of a funnel (the only allowance for error in the whole scenario), roll around the edge a few times, and fall through. As it came out of the funnel, the cannonball would hit against a lever that started a circular saw. The saw would cut through a rope, which was holding up a telephone pole. Slowly the telephone pole would begin to fall, and too late Foghorn Leghorn would realize that the fascinating show was at his expense. As he turns to run, the very tip of the telephone pole smacks him on the head and drives him like a peg into the ground."¹

Behe goes on to say that this Foghorn Leghorn cartoon contraption, as a Rube Goldberg-like machine, is "irreducibly complex." This of course means that if any one part is removed, the whole machine fails and the desired end result or function, does not occur. Behe compares such thought contraptions to real life systems of functions within living things, such as the clotting cascade in the blood coagulation pathway. This clotting cascade works in just about the same way that Foghorn Leghorn was whacked into the ground by the telephone pole. Each event in the clotting cascade must happen before the next event can occur. If any one event is blocked, clotting will not occur at all. Behe wonders how such a system could have evolved by mindless naturalistic mechanisms?

Well, we must first ask if Rube Goldberg machines really are "irreducibly complex"? It is true that if one part is removed the cascade is interrupted at that point. It is kind of like taking a domino out of a line of dominos that are balanced on their ends. The dominos after this point will not be knocked over by the previously cascading dominos. In this way, the final function of the cascade is rendered useless. However, the cascade is in fact still able to function from the point of interference onwards. So, its function cannot be said to be completely destroyed just because of a single point of interruption. Also, by the very nature of cascades, more parts can be added on the originating end of the cascade to make it longer and longer still... like adding more and more dominos onto the end of a chain of dominos. At first one starts with one domino, then two, three, four and so on until one has a very long cascade set up.

Consider the Foghorn Leghorn cascade again from this perspective. What if the entire cascade describing Foghorn Leghorn’s demise started simply? Foghorn picks up the dollar that is attached directly to the telephone pole by a string. When the string is pulled, the carefully balanced telephone pole falls over and drives Foghorn into the ground like a tent peg. Granted, this certainly is not nearly as interesting or entertaining. But, it would work - right? Now, what if we add just one little part to the cascade? Lets add the rope that holds up the telephone pole and a saw that cuts the rope. The string is attached to the switch on the saw. When pulled, the string turns the saw on and it cuts the rope and the pole falls. A bit more interesting and it still works. Now lets add one more little part. Lets add the cannon. The string pulls a match and ignites the cannon and the cannon ball hits the saw switch which cuts the rope that holds the pole… and now we are getting a lot more interesting! We are evolving a complex cascade one small part at a time - right? It sure looks that way.

This argument is in fact commonly used as an explanation for the origins of such apparently complex cascades as occur in blood clotting systems, visions systems, and energy metabolism systems to name just a few. After all, according to the theory of evolution, very large and apparently impossible tasks are broken up into manageable parts. This is Dawkins's main argument in his book, Climbing Mount Improbable. An impossible statistical cliff that cannot be scaled in a single bound by natural selection is scaled in small little evolutionary steps. Actual laboratory experiments have been put forward to support such a concept. What is especially interesting is that these laboratory experiments actually work! It has been demonstrated in real time that the addition of unique components onto the end of a metabolic cascade is in fact possible.

The evolutionary biologist Kenneth Miller described one such experiment in his recent book, Finding Darwin’s God. Miller quotes an experiment done in 1982 by professor Barry Hall.² In this experiment Hall deleted part of a metabolic pathway in an E. coli bacterium. Using the proper environmental pressures, the bacterium "evolved" the missing piece of its cascade back again!³

This missing piece was actually the tip of a long cascade that breaks down sugar molecules and extracts energy from them. Glucose is the main sugar utilized by this cascading pathway. There are of course other simple sugars, such as galactose and fructose etc., that can enter this pathway as well . Each of these also requires a unique enzyme or enzyme pathway to convert them to something that can enter the cascade. Then, just like adding more events to Foghorn’s cascade, more events can be added to the sugar cascade. You see, there are different kinds of sugars. Some of these sugars are more complex than glucose, but can be broken down into glucose and/or one of the other more simple sugars that are already part of the existing cascade. Once this breakdown occurs, a complex sugar molecule becomes just another part of an extended sugar metabolism cascade. The problem is that unique enzymes are needed to break down complicated sugar molecules. Some sugars may even need more than one unique enzyme to break them down to a point where they can enter the established cascade. However, the benefits of obtaining these enzymes are great. If such cascades of complex sugar breakdown can be established, any bacterium with such capabilities would survive better than its peers in an environment that was rich in a particular complex sugar.

The ability to evolve such advantageous enzymes would certainly enhance the survival of the species. In fact, the "evolved" bacteria in Hall's experiments quickly outgrew those that had not yet evolved the needed enzyme. Of course, this is only natural. It is the law of survival commonly known as "the survival of the fittest". Hall went on to demonstrate the evolution of two and even three additional steps added on to the original sugar cascade.

Although I do agree with Behe when he says that cascades are indeed irreducibly complex, I do not agree with his assertions that all such cascades are theoretically or even practically impossible to evolve via the mindless processes of random mutation and natural selection. Cascades in living systems are certainly as complicated and even vastly more complicated than the one that whacked Foghorn Leghorn, but even such complexity does not seem to be completely out of reach in certain cases.

As already described, the removal of one part of a cascade may not destroy its ability to perform. The removal of an enzyme that allows the utilization of complex sugars does not eliminate this cascade's ability to continue to break down glucose or galactose or fructose. Even the removal of the enzyme needed to break down glucose itself is not vital to the function of the rest of the cascade. Even though glucose can no longer be utilized, fructose still can be, along with several other types of sugars. A cascade is therefore reducible without the loss of all function, but it is still "irreducible" as far as the function that it just lost is concerned. For example, a minimum number of parts are needed that have a fairly specific internal structure in order for a bacterium to be able to utilize the lactose sugar for energy. This minimum part requirement creates a degree of irreducible complexity. Not just any series of parts will do. Specific protein "parts" are needed. In fact, Hall's experiment illustrates this specificity very well.

Hall did delete the gene needed to produce an enzyme (lactase) that broke down the sugar lactose into two other sugars called glucose and galactose. Both glucose and galactose are part of the established sugar metabolism cascade of E. coli. So obviously, without lactase, an E. coli bacterium can no longer utilize lactose for energy. Hall deleted that lactase genes to see if the E. coli bacteria would "evolve" back the ability to utilize lactose when grown on a lactose enriched media. Hall's experiments were a stunning and dramatic success. His colonies of E. coli quickly "evolved" the ability to use lactose. There is just one little catch. Hall did not delete a spare tire gene (the evolved beta-galactosidase gene - ebg) that required just one point mutation to produce an enzyme with a fairly high level of lactase activity. But what if the E. coli had not been so fortunate as to have this spare tire gene? What would have happened then? Hall wondered about this himself. He then deleted the spare tire gene as well as the original lacZ genes. Would there be lactase evolution now? So far, none of these large populations with high mutation rates has ever evolved the lactase ability despite being subjected to lactose enriched selective media for tens of thousands of generations. Evidently, no other gene could lend its information by itself or in combination with any other gene to aid in the bacterial "evolution" of lactase. Hall described these particular bacterial colonies as having "limited evolutionary potential."³ But why did these colonies have limited evolutionary potential?

It turns out that there are statistical gaps that separate unique protein/enzymatic functions from each other. Not every protein sequence will be recognized as "beneficial" by a given bacterium. In fact, the vast majority of possible protein sequences required for the function of a particular enzymatic activity, like the lactase function, will not be able to produce this specified function - even a little bit. Because of this problem, if proteins are not already very very close in sequencing to begin with, the statistical odds that one will simply "evolve" into another are remote because they are separated by a vast number of "neutral" amino acid sequences. Neutral proteins cannot be guided by natural selection along any evolutionary path whatsoever. Why? Because nature only sees functional differences. Nature cannot guide if it is blinded by functionally neutral differences. Thus, nature cannot guide evolution across neutral gaps. Obviously then, without this guidance of natural selection, evolution simply stalls out after an endless "random walk" and dies.

The results of Hall's experiment certainly prove Behe's point that at each step in a cascade the required protein-enzyme is, in itself, irreducibly complex to at least some degree. It also proves that this level of complexity is just plain out of reach for many kinds of bacteria (not just the double mutant E. coli studied by Hall). A minimum number of amino acids are needed in a specific arrangement in order for a specified function to be realized. If such an enzyme is altered or reduced beyond a certain point, the specified function will simple vanish completely. The genome that is left may not have what it takes to cross the resulting neutral gap in function between what is and what is needed. It is this neutral gap in function that forms the basis of Behe's argument that irreducible complexity poses a significant hurdle for the theory of evolution.

Of course, a cascade is no more complicated than the most complicated link in its chain. If this most complicated link can be overcome, then the rest of the chain would be easy to make. The question is, can the most complicated link be overcome in a reasonable amount of time? Professor Hall showed how simple bacteria can sometimes evolve links in a cascade chain (but not always). If these links were all insignificant hurdles they could simply be added up to produce something quite significant - like a blood clotting cascade or a metabolic pathway. But, as previously described, there might be a few snags along the way if the links themselves are simply too irreducibly complex, in and of themselves, to be overcome by a given life form in a given environment - even in 4 billion years.

Professor Hall never evolved anything that crossed a neutral gap that was more than two mutations wide. The single non-functional gap of two mutations that he did cross, he could not explain. In fact, by his own calculations, he figured this feat to be impossible - taking an average of 100,000 years to cross. The apparent success of the crossing of even this tiny gap of non-function astounded him. He attempted to explain the success of this crossing by saying, "under some conditions spontaneous mutations are not independent events."³ He went on to say that this is, "heresy, I am aware." If it is difficult for professor Hall to imagine the crossing of such a small neutral gap, what would it take to cross a three mutation gap… or a four mutation gap?

As it turns out, Hall was mistaken in his calculations. He based his estimates on a mistaken understanding of how the statistics of random walk works. Hall assumed that each new mutation along the road toward the desired beneficial mutation would have to become "fixed" or transferred to the entire population before the next mutation could arrive in an additive way. To understand how this need not be the case, consider the following hypothetical situation:

Starting with a steady state population of a trillion trillion (10²⁴) random sequences of 200 amino acids, each mutating every second into a new sequence, how long would it take, on average, to find any one of the 10e200 lactose cleaving proteins? Surprisingly, it would take around 500 trillion trillion years to find even one of the 10e200 lactase sequences on average. In order to solve this problem, one must first calculate how many different possible proteins could be made with a series of 200 amino acids. The answer is on the order of 10e260 (1 with 261 zeros following). Compared to this number, even an astronomically large number like 10e200 is relatively minute. What happens is that each of the 10e200 functional proteins is surrounded by trillions of non-functional protein sequences.

The "potential space" of 10e260 options can be visualized as a gigantic chess board. Each square on the chess board represents a different amino acid sequence. Each member of a population can only occupy one square at a time. A limited population simply cannot cover all the potential squares on the chess board at any given moment of time. With each mutation to an individual, it changes squares. If any one individual comes across a beneficial sequence, like the lactase enzyme sequence, that individual and its offspring will tend to stay on that square because of the selective advantage given by that square. This advantage will be translated into an increased population on and immediately around that particular square of the chess board.

The interesting thing about random walk is that with each doubling of the length of the average random walk, the time involved increases by a factor of 2. For example, if the average random walk required for a particular colony of bacteria to achieve a particular level of complexity required 5 neutral steps or changes in DNA, the total number of options or potential space on our chess board would be 4 to the power of 5 or 1,024 squares. Depending on our population's size and mutation rate, we could estimate an average time required to reach all of these squares at least once beginning at a random starting point. The bigger our population, the faster we could reach all the squares.

For instance, if we started out with a population of 1 trillion bacteria and if all of these bacteria started out on one square on our chess board, it would take around 65,000 generations for them to reach equilibrium over all the squares of the chess board. At equilibrium, about 0.098% of them will be on each one of the squares of the chess board. Even though 0.098% doesn't seem like a big number, it actually works out to be 9.8 billion out of a population of 1 trillion. In other words, after 65,000 generations, there would be an average of 9.8 billion bacteria covering each of the 1,024 squares on our chess board of potential space. So obviously a gap of 5 neutral mutations would not be a problem for a population of 1 trillion bacteria to cross in relatively short order. But, what happens if we double the gap to 10? A gap of 10 mutations/steps would create a chess board with over 1 million squares of potential space (1,048,576 to be exact). At equilibrium, our population of 1 trillion would have only 953,674 individuals on each of the squares instead of the 98 billion it had when the gap averaged only 5 steps wide. Doubling the gap again to 20 steps makes our chess board grow a million fold to just over 1 trillion squares of potential space (1,099,511,627,776). Now, our population of 1 trillion would average a bit less than one member of the population on any one square at any given point in time. I think the trend is obvious by now, but just for kicks, doubling the gap again to 40 steps increases the size of our chess board a trillion fold to just over 1 trillion trillion squares. Now, each of the members of our population of one trillion are surrounded, on average, by one trillion empty squares that they have to search out all by themselves.

On average then, each one of the members in our population will have to experience over 1 trillion mutations (with none of them repeating) in order to reach all of the potential spaces on our chess board. The time required for this process truly reaches astronomical proportions in short order. With each doubling of the neutral gap, the average time required increases by a factor of two. The only way to reduce the required time is to increase the population's size by a factor of two. This does help for a while, but very quickly the required size of the population needed to keep up becomes impractical for any environment to support and further evolution simply stalls out.

Consider now that many enzymes/proteins are hundreds or even thousands of amino acids in length - and very specified in internal structure. The problem seems clear. In fact, because of this problem all living things may have very "limited evolutionary potential" indeed! Bacterial colonies such as Salmonella, Proteus, and Pseudomonas as well as Halls double mutant E. coli colonies can be grown on Hall’s selective media or any selective media in any sequence until the cows come home and none of them will ever evolve the lactase enzyme. If this relatively simple function does not evolve in certain creatures with "limited evolutionary potential" what about functions that require multiple proteins or systems working together simultaneously? How would such a multi-protein function evolve gradually? For example, bacterial motility can come in many different forms to include flagella, cilia, undulating membranes etc. However, all known motility systems require many protein "parts" working together simultaneously in a specified arrangement with each other (much like the specified arrangement of amino acids required by single-protein enzymes to work - only on a much larger scale of complexity). If it is often difficult for various life forms to evolve single protein enzymes with relatively few specified amino acids, imagine the difficulty required to evolve a multi-protein function where entire proteins are specified in their arrangement with each other. In fact, the odds are so bad that even on paper the evolution of such a multi-protein function is simply beyond any reasonable explanation - not to mention the fact that such a multi-protein system of function has never been observed to evolve in real time in either the laboratory or in nature.

For example, an electric motor is irreducibly complex and it is not a cascading system. It has a minimum number of parts working together at the same time, in a specified orientation, in order for it to have a function. It needs a wire coil, a surrounding magnet, electrical current and bushings to switch the electrical current back and forth at the proper moment in time. All these parts must of course be set up in the proper relation to each other. However, if one part is taken away, none of the other parts will "work" together - period. There is no "cascade" of function since all the parts work together at the same time. So, in order to get any function whatsoever from the electric motor, all the needed parts must come together in a highly ordered/specified way - suddenly. An electric motor, minus one of its parts, has no function. It might as well be sitting in a junk pile at the city dump.

What is especially interesting is that some bacteria, as Behe points out, have little mechanical motors that are very similar to electric motors. Just like mechanical motors they require a minimal number of protein parts (each of which are in turn made up of hundreds or even thousands of amino acid "parts" in specified order), all working together at the same time, in a specified orientation, to produce their rotary motion. If one adds up all the required amino acids as "parts", literally hundreds of thousands of parts are required in a specified order before function can be realized.

Of course, we are talking here about the famous bacterial flagellum illustrated above. Flagella are long whip-like cords attached to certain bacteria by a little motor of sorts.¹ The motor actually spins and causes the flagella to spin. The spinning flagella propels the bacterium through the liquid that it is swimming in. The problem here is that the motor that spins the flagella is not only "irreducibly complex", but has a very high level of specified complexity (a large number of required parts all working together at the same time). It is made up of a fair number of fully formed specified proteins (50 or 60 different ones), which are in turn made up of hundreds or thousands of specified amino acids. Many of these parts are irreducible. If even one of these irreducible parts is removed. The flagellar motor will not work, not even a little bit.

Each of these parts is, in itself, complex - just as each link in the cascade system is complex. However, what separates this type of irreducibly complex system from a cascading system is that each of the parts is required to be in a specific orientation with all the other parts. This adds another specified constraint to the system, which raises the level of functional complexity significantly. In other words, a lot more information is required to code for a flagellum than for a cascade of enzymes of equal size and number. This creates an even larger neutral gap between functions at this level of complexity as compared with single protein enzymes or cascading systems. Again, no such multi-protein function has ever been seen to evolve in real time. It just seems to be too far beyond what the mindless processes of evolution are capable of - even with trillions and zillions of years of time.

Some do try and explain flagellar evolution by proposing that many of the parts in a flagella are used as parts in other cellular systems of function. For example, the actual flagella is very similar in structure to long tube-like secretory structures. In fact, it is thought that some flagella might function as both a motility structure as well as a secretory structure. Likewise, many of the other parts in a flagellar motility system do other similar jobs in other systems within the cell. Therefore, it seems obvious to many that all these various parts already existed and therefore might easily come together to form the motility system diagrammed above. The problem is that the parts do not naturally self-assemble into a flagellar system, or any other system of function. They must be directed in their assembly by a pre-formed coded message contained in that cell's DNA. It is like the parts of a car. Many of the same parts could be used to build a boat, or a house, or any number of other things. However, planning is needed because the parts themselves, if left to themselves, do not self-assemble to form any one of these things (even if given plenty of energy and opportunity to interact with each other). Likewise, starting with a bacterium without a motility system but with all the needed parts to make one, there is no series of point mutations that will get it from what it has to the goal of motility without the crossing of neutral gaps in function (despite the selection pressure or advantages of motility should this function happen to evolve). In other words, there is no way to mutate the genetic code were each and every mutation will be beneficially advantageous until motility is achieved.

So, Miller’s attempt to dismiss Behe’s challenge of irreducible complexity stands on very shaky ground. At best it is limited to the level of single protein enzymes and relatively simple cascading systems of single protein enzymes. Even these examples fall very short as cascading systems are far less complex than multi-protein systems where all the protein parts work together at the same time in a specified arrangement. In many cases, even these lower level cascading functions pose significant hurdles that severely limit the mindless processes of naturalistic evolution. Obviously then, it is only intuitive that anything of even slightly greater complexity might just stall out the processes of evolution completely. And, in real life, this is exactly what we observe. Again, as far as I am aware, no life form in any environment or series of environments has ever been shown to evolve any multi-protein system of function where the proteins all work together at the same time in a specific orientation with each other.

So, even though some irreducible functions can evolve and have evolved, there is a ladder of irreducible complexity that quickly moves beyond what evolution has done and can do. In my opinion then, Behe's concept of irreducible complexity remains pretty much untouched as a challenge to the modern theory of evolution and as a strong voice in favor of intelligent design.

Behe, Michael J. Darwin's Black Box, The Free Press, 1996.
Miller, Kenneth R., Finding Darwin's God, HarperCollins Publishers, 1999.
B.G. Hall, Evolution on a Petri Dish. The Evolved B-Galactosidase System as a Model for Studying Acquisitive Evolution in the Laboratory, Evolutionary Biology, 15(1982): 85-150.