PHYSICS TODAY / March 1982, p. 26
In the Guest Comment in April '81 (page 9), Frank Tipler argues that if extraterrestrial beings existed, our galaxy would be so full of their Von Neumann machine proxies that we could not have missed them. Tipler replaces Michael Hart's and James Trefil's humanoid colonizers with machines, to circumvent the costs of humanoid support and humanoid fragility. Francis Crick and Leslie Orgel replace them with cargoes of microorganisms in a revival of Svante Arrhenius' century-old "Panspermia" hypothesis.1
I'm concerned that rebuttals of Tipler's argument may refuel the push for a Search for Extraterrestrial Intelligence (SETI) which has been steadily championed by a group of physical scientists led by Frank Drake, Carl Sagan, Bernard Oliver, John Billingham and Philip Morrison. Therefore, I would like to offer less imaginative, but perhaps more compelling, biological arguments against SETI.
Let me say at the start why I believe many scientists are so easily gulled by the intriguing idea of ETI: It's that they simply have faith that the deterministic laws of chemistry and physics assure that all classes of macroscopic processes, including those of biological evolution, must be repeated countless times, over the multi-billion-year lives of the galaxies and the vast stretches of the universe. However, such ergodic faith is probably mistaken. Macroscopic Darwinian selection is unparalleled by other physical processes and is much less likely to repeat itself. It tests the environmental fitness of new genetic messages (mutations) that, prior to selection, have been generated by microscopic random accidental modifications of pre-existing molecular messages. These molecular messages appear to be otherwise archivally insulated from environmental editing. The full set of messages tested by selection, from the beginning of life, constitute only a minute and probably unrepresentative sample of different possible messages from which the sample has been "drawn". The number of possible messages exceeds the estimated number of atomic particles in the entire universe by some 100 million orders of magnitude!2 Therefore, no matter how prevalent life might turn out to be, biological evolution on earth can easily have generated many "inventions," perhaps including intelligence, which are unique in the universe.
But let's examine the main arguments put forward by the proponents of SETI and see how convincing they are. Drake has said, "At this very minute, with almost absolute certainty, radio waves sent forth by other intelligent civilizations are falling on Earth" (in Ferris,3 my emphasis). Drake has developed an equation to estimate, as he puts it, the number of communicative civilizations we might find in our Galaxy.
N = R*fpnpflfiftL
(To my knowledge, Drake has not published this equation, and so it appears by attribution, as in Tipler's article, or as a quotation.3)
R* is the rate of star formation, averaged over the lifetime of the Galaxy, in units of numbers of stars per year;
fp is the fraction of the stars which have a planetary system; and
np is the mean number of planets within such planetary systems which are ecologically suitable for life.
Drake, quoted in Ferris,3 informs us:
" fl is the fraction of potentially life-bearing planets that actually do give rise to life. Biochemical experiments on Earth suggest that this number is approximately 1, that is, that the chemistry of life is ubiquitous and powerful, and that life tends to arise wherever it has a chance to do so."
" fi is the fraction of living systems that evolve intelligence. The fossil evidence on Earth suggests strongly that this number is also 1. We say that because in the fossil record, there is one category of things that constantly improved and that is brain size, which we associate with intelligence."
ft is the fraction of those planets on which intelligence has arisen, which then develop the techniques of radio communication.
L is the average life-time of these hypothetical civilizations.
Drake estimates the first six variables as 1 per year, 1/2, 2, 1, 1, and 1. respectively. He argues that the seventh, L, is the only variable with uncertainties of large magnitude.3 If L lies between 10 million years and 10 years, our radiotelescopic search must provide sufficient sensitivity to detect signals coming from sources that are, respectively, between about 100 and 10,000 lightyears away to provide a reasonable probability of detecting one such signal source.3 If the searchers for ETI are lucky, and L = 10,000 years or more, then N = 10,000 or more communicative civilizations in our Galaxy, and a proposed 14-million-dollar SETl effort has just a bare chance of success. It would search our galactic neighborhood out to about 600 lightyears (between 10-4 and 10-5 times the volume of the Galaxy) with significant sensitivity at that distance only for signals 100 times stronger than our strongest radio signals (those produced by our own ballistic missile early warning system radars).
While estimates of R*, fp, np, and ft may contribute more uncertainty than Drake appears willing to admit, I'll show that estimates of fl and fi are also extremely uncertain, severely diminishing confidence in the chance of success. Estimates of fl and fi must depend upon what we know of biology, rather than astronomy; and perhaps even more importantly, upon the rules of inference concerning how confidence or degree of certainty is rneasured.
Let's look at the case for fl.
It's a fact that life evolved on Earth. But did it originate more than once? All known organisms use the same genetic code and coding mechanisms, and it therefore seems extremely likely (though not absolutely certain) that all are descendants of a common ancestral "cell." This, by itself, could be taken as evidence that life arose but once on Earth. Some argue, as an element of faith in macroscopic determinism (as Drake does above), that the nature of the chemical processes on an Earth-like primordial planet probably guarantees that life will evolve with such a code and coding mechanism. And, perhaps, that if life arose a number of times on Earth, all organisms would nonetheless seem closely related, even though they arose independently.
But nearly all agree that, averaged over the planet, the primordial "soups" in which life could arise would have to contain nearly equal numbers ("a racemic mixture") of levo (L) and dextro (D) versions of whatever asymmetric organic molecules served as building-blocks for the first organisms. Molecular genetic coding mcchanisms work by the precise fitting together of molecular pieces on molecular templates. Right. hands don't fit well into left-handed gloves. Therefore, one of the first molecular accidents that produces a primitive organism could lead to an ability to use only L or D isomers as it's building blocks. If life arises frequently, roughly equal numbers of organisms will appear which use either the L or D components of the primordial soup.
However, as Louis Pasteur discovered, all known organisms, from bacteria to men, select only L amino scids to construct their proteins. This otherwise still inadequately explained violation of the law of conservation of parity provides an independent argument, first considered by J. B. S. Haldane in l929, that life arose only once on Earth. By extrapolation, spontaneous generation of life may be an improbable event on any Earth-like planet. Of course, it might be argued that the first organism and its descendants almost immediately made the thousands or millions of pools of soup on this planet uninhabitable for later-arising forms of life using D amino acids, but this, in itself is highly improbable. The first orgsnism was undoubtedly an inefficient sort compared, sav, to a modern bacterium. It had a very limited repertoire of biochemical pathways, and most probably had a negligible impact on those components of the soup it could not directly use. And until it evolved some protected, spore-like, germinal stage, it couldn't even spread to uninhabited pools, except under the most extraordinary circumstances.
Therefore, for some time, L and D forms would have occupied somewhat isolated biochemical niches, and some evolutionary diversification of both lines should have occurred before they came into competition. When new evolutionary inventions appear in separate lines occupying separate niches, they don't easily lead to the extinction of one of the diversified lines (even in predator-prey relationships), as witnessed by the persistence of procaryotes after eucaryotes evolved, of one-celled organisms after multicellular types had appeared, and so forth. Therefore, the evidence all supports the proposition that life arose only once on Earth.
The confidence limits associated with a predicted recurrence of an observed or inferred event narrow as we increase the precision with which we can characterize a probahility difitribution for that class of events. But if only one event can be inferred. and if we have no separate way for estimating properties of such a distribution (for example, by extrapolation from observations of another class of events that we believe is a good model), then all guesses about the probability of recurrence are worthless.
Sagan and Drake say, "Since life originated on Earth in a span much shorter than the present age of the Earth, we have additional evidence that the origin of life has a high probability, at lesst on planets with an abundant supply of hydrogen-rich gases, liquid water, and sources of energy."4 If we were to believe that worldwide conditions were suitable for the unique event to have occurred only during a short interval, this would allow, but not require, a high probability (or equally, a low probability) of occurrence per unit time during the short interval. Therefore Sagan and Drake's position is indefensible. The primordial conditions terminated within about 600 million to 1 billion years after the origin of life because the descendants of the first organisms gradually consumed the primordial soups, and because an oxygen-rich atmosphere began to develop after the invention of aerobic photosynthesis. If life had not apeared, it seems likely that more or less primordial conditions would have persisted to the present time. However, this provides no information for modeling the probability distribution in a way that significantly increases the worth of any guess for fl.
To recapitulate, using the classic model for drawing inferences from random samples: If we have reached only once into a bag of marbles (representing the Earth-like planets in our Galaxy) and have withdrawn only one marble, and that marble is inscribed with a particular message set, written with the genetic code (representing the observation that life, the genetic coding and copying mechanism, has arisen on Earth), then we have just as much right a priori to guess that the bag contains only one-in-a-million (or even no) other marbles with similar inscriptions, as we have right to guess that all are so inscribed. Yet when Drake sets fl = 1, he is guessing that virtually all marbles are inscribed with similarly coded messages. Therefore, we should have little confidence in his guess. It can't support anything remotely approaching "almost absolute certainty." The L amino-acid story alone might better support a guess like fl = 10-6.
Now, let's look at fi
It's a fact that one line of organisms has evolved on Earth that is intelligent enough to develop radio technology. It's not at all clear that gorillas, porpoises, or dogs would ever give rise to lines that would be capable of developing such technology (even though gorillas have now been taught to use large vocabularies). However, let's suppose they would. Since they and man are all mammals, I argue that they represent a single origin of intelligence. Therefore, to guess from this that fi = 1, is still another example of trying to predict the characteristics of the contents of the bag from having seen only one marble. Can we infer anything useful about the probability of fi from what we know of the evolutionary record and evolutionary processes? Looking back in evolution, before the development of the vertebrates, we can recognize inventions, each of which appears to have been a sine qua non for the evolution of intelligence. A reasonable (but not exhaustive) list includes evolution of an eye, a nerve cell, multicellularity, the cilium/flagellum and the eucaryote way of life. This suggests representing fi as a product of probabilities, fi = fiefinfimfic fiu. Except for the cases of eyes and multicellularity, the record indicates that each of the other innovations occurred only once on Earth. There is simply nothing remotely approaching a precursor of intelligence in any lines that lack these innovations, and very little more in the lines that contain all of them but eyes. Therefore, even if the values for fim and fie are appreciable, fi still could be vanishingly small. If fi = l, this is equivalent to saying that what is called evolutionary convergence has a very high probability of occurrence, even at very high levels of complexity. 2 Therefore, we should look for a model for the probability distribution for fi . We might attempt this by trying to estimate the probability that one or more distinct and ancient lines of organisms (squid-like? sea-squirt-like? ant-like? venus-fly-trap-like?) would finally deveiop an equivalent level of intelligence if the vertebrate line had been snuffed out near the time of its origin (early in the Silurian Period, about 400 million years ago, which is between 2% - 4% of the age of the universe). The squid, sea squirt, and ant are at least plausible candidates because they and man have nervous systems. They evolved from a common ancestor with primitive nerve cells, and are therefore part way along a "right" path. (And the sea squirt is a close relative of the vertebrates.) Organisms which don't share this common ancestor (such as, the plants and fungi) lack even primitive analogs of brains, and are much less plausible candidates. But a very large number of extremely improbable events would be required for "known" evolutionary processes to remodel the nervous systems and lifestyles of even squid-like, sea-squirt-like, or ant-like species to converge functionally towards that of man. In fact, there is no precedent in the living or fossil record for evolutionary convergence of such a magnitude. On these grounds, one should suspect that fi is much smaller than 1.
The synthetic or neo-Darvinian interpretation presents evolution as an esisentially random but "opportunistic" process for producing adaptive mechanisms. Most observations support this model. According to the "central dogma" of molecular biology, mutations, duplications, deletions and rearrangements occur as errors in the copying of genetic DNA -- the blueprints and instruction manuals of organisms. In the very rare cases where the reading of such errors yields improved competitive performance, the carriers of the new genes then usually replace the old model in the environmental marketplace at rates which usually vary inversely with the size of the breeding population. That is, evolution "takes advantage" of the opportunity provided by a molecular-scale accident. The probability of producing convergence in different evolutionary lines by this process is rapidly diminished, the more complex the adaptive mechanism and the further apart are the origins of those lines. If this is the whole story, fi is probably infinitesimal. One may however wonder, is evolution sometimes also propelled or guided unerringly (in some as yet unspecifiable, but well-determined way -- perhaps through the Lamarkian inheritance of acquired characters) to certain specific goals? Can evolution easily "rediscover" certain useful classes of "technical solutions" to problems of survival, more or less independently of evolutionary starting points, and with much higher frequency than would be estimated by an opportunistic neo-Darwinian analysis? The formulation by Sagan and Drake of this hypothesis is, "There might be a kind of biological law that there are many paths to intelligence and high technology, and that every inhabited planet, if it is given enough time and does not destroy itself, will arrive at a similar result."4 Is there any evidence for such a highly deterministic law?
The issue of multiple independent origins of intelligence has not been a preoccupation of many serious evolutionists (Drake's non sequitur comment about the relationship of brain size to intelligence notwithstanding). But evolutionary convergence of other biological adaptations has been of considerable interest in comparative biology. A classic case of convergence, which continues to elicit attention, is the independent evolutionary origin of flight mechanisms among insects, fish, reptiles, birds, and mammals. By noting how often various classes of technical solutions to problems of survival, at different levels of complexity of mechanism, and in lines of varying degrees of relatedness, have evolved independently on Earth, one might be able to extrapolate a distribution function for independent origins of high-level intelligence. In effect, we might be able to test the hypothesis of highly deterministic (as opposed to opportunistic) convergence and perhaps develop a basis for estimating fi .
The repeated independent development of flight -- or of fish-like streamlining for aquatic life-forms -- represent convergence at relatively low levels of complexity. Opportunist convergence is compatible with these cases. The best known cases for apparent independent evolutionary convergence of significantly more complex mechanisms in relatively distantly related lines involve some 15 types of camera eyes distributed among the several animal phyla.5,6
The prior development of a complex camera, with its associated pre-processing circuitry, is probably essential for the further evolution of a nervous system that could permit an organism to develop high-level intelligence. Therefore a discussion of the convergence of camera eyes is especially relevant in an attempt to estimate the probability of independent origins of intelligence. The cephalopod and vertebrate eyes constitute the case, par excellence.
Comparison of generalized fish eye (left) and generalized cephalopod eye (right). The thickness of the retinal layers (typically 200 - 300 microns) has been exaggerated to make the differences between the inverted (left) and verted (right) retinal structures more apparent. In cephalopods (right) the lens forms by the fusion of two epithelial layers, giving rise to two distinguishable parts; in the squid and cuttlefish, they form an essentially single simple spherical lens (as illustrated here), while in the octopus the two portions often have appreciably different refractive powers. (Click here to download a higher resolution copy of this figure.)
The overall designs of these two cameras are strikingly similar (see the figure). For example, an external cornea provides a fixed portion of the refractive power of the lensing system (the major portion for land animals). A muscularly controlled variable iris diaphragm adjusts the average level of illumination in the image plane. An internal lens, muscularly deformable or movable (depending upon species) provides variable focus (and virtually all of the refractive power in aquatic species). Behind and between the photosensitive elernents of the retina are darkly pigmented portions of a light-trapping layer of cells which function like the blackened interior of a camera.
The vertebrate eye is characterized by a so-called "inverted" retina. The layers of nerve cells of the retina lie in front of the light-sensitive rods and cones from which they receive excitatory or inhibitory signals (that is, the nerve cells lie in the light path), and the nerve-cell axons exit along the surface of the outer layer of nerve-cell bodies and converge towards the center of the retina where they bundle together to form the optic nerve, which then passes through the retina out the back of the eye and to the brain. This construction imposes special restrictions on the material composition of those nervous layers. The nervous parts of retinas are functionally the most complex parts of the eye, acting as arrays of parallel computers to preprocess images before passing information on to the brain. In the vertebrate construction, they must also be almost optically isotropic and transparent so as not to degrade the image formed by the light that must pass through them. In contrast, the upright or "verted" retina of a typical cephalopod (for instance, squid, cuttlefish or octopus) avoids this problem quite simply. The nervous layers lie out of the way, behind the light-sensitive layer in the optic ganglion in a more "rational" design.5
The light sensors (the rods and cones) in the vertebrate retina are clearly assignable to the class of "ciliary" receptors which appear to trace their evolutionary origins to the light-sensitive cell membranes of flagellae and cilia adjacent to the "eye spots" of primitive protozoa and algae (similar to the modern unicellular alga, Euglena).6 On the other hand, the light sensors of the cephalopod retina, "rhabdomeric" receptors, were, for a while, presumed to derive from very different membranes, but it now seems likely that they also are derived from membranes at the bases of cilia.6
Vertebrates are members of the phylum Chordata. The living protochordates, such as Amphioxus, acorn worms and sea-squirts, have extremely primitive light-sensing organs, hardly qualifying as "eyes," let alone camera eyes. The presence of strikingly similar camera designs in the vertebrates and cephalopods, with apparently large divergence in the structure of the receptor and data-processing layers -- combined with an apparent lack of anything but the most rudimentary light detectors among the living protochordates, all seem to provide compelling evidence for complex convergence in independent lines that first diverged from one another from a common eyeless ancestor. But maybe not.
It is well known that populations of organisms, isolated in lightless environments -- typically in caves -- tend to evolve rapidly to eyeless forms. Species of blind cave fish, salamanders. and arthropods are not at all uncommon. They illustrate evolutionary simplification by loss of function. This pattern of evolution suggests that in environments where sight has little or no value, the eye structures burden their carriers with functional or energetic costs that become overwhelming in competition with mutant eyeless forms of the same species. Simplification by loss may, in fact, be "sufficient to account for the apparent absence of photoreceptors in the ancestral types of so many phyletic lines,"6 especially if due to deletions of parts of chromosomes, because they are essentially irreversible copying errors.
So there may be more to the story of camera eye convergence than meets the casual vertebrate eye. What appears like pure convergent evolution might represent parallel evolution from a common ancestor with a camera eye, obscured because the specialized ancestors of living ''primitive'' members of either (or both) lines have experienced considerable evolutionary simplification by loss of function. In that case, many fewer improbable random errors would be required to produce the observed similarities between vertebrate and cephalopod eyes.
There is a well-known biogenetic law, first proposed in 1866 by Ernst Haeckel, "ontogeny recapitulates phylogeny." This statement summarizes an enormous range of observations of embryonic development throughout the animal kingdom. Embryos typically develop a sequence of structures which reflect the order of their appearance in their ancestral lines, even though these structures may be transformed or lost further along in embryological development. By such criteria, the phylogenetic origins of the vertebrate eye are very ancient. In the vertebrates, the eye begins its appearance with, or before, the notochord and gill slits, which are two of our three main ties (dorsal nerve chord, notochord, and gill slits) to the living protochordates. This provides another reason to suspect that our earliest chordate ancestors had camera eyes instead of rudimentary photoreceptors like those of the protochordate, Amphioxus, that lives mostly buried in the sand. Because eyes are composed of tissues that don't fossilize, the fossil record can't help us to resolve the question of whether living protochordate visual systems represent examples typical of ancestral chordates or, alternatively, illustrate evolution by loss of function .
What's known about the evolution of the cephalopod line? The cephalopods are the most advanced members of the phylum Mollusca. A wide variety of eyes are found among more primitive molluscs.5,6 The eyes of some bivalves illustrate a series of transitional forms5 that "start" with a primitive verted camera eye with rhabdomeric receptors and "end" with the Scallop's eye, a Schmidt-type camera with catadioptric (reflecting-refracting) optics7 with an inverted retina with ciliary receptors. In man-made Schmidt cameras, the detector assembly (or a mirror representing it) obscures part of the incoming beam of light. The Scallop's inverted retina also faces a reflecting hemisphere to the rear. In such a design, transparent nervous layers minimize that obscuration. This may reveal the rationale for the origin of inverted retinas. Later conversion back to fully dioptric camera design, as in, for example, the vertebrate line, would leave the inverted retina as a vestigial structure, much more closely related to the verted retina than has previously been suspected.
The coelenterates constitute a phylum which is considered to be more ancient than the molluscs or chordates, and might have contained the ancestors of both. Among the coelenterates are some cubomedusan jelly-fish that have simple, verted camera eyes.8 Is it possible that an ancestor common to the cephalopods and the vertebrates had eyes similar to the cubomedusan? If that were the case, the observed convergence would hardly be so remarkable; probably much less so than what was surely independent evolution of flight in the vertebrates and the insects. This hypothesis is now testable by protein and nucleic acid sequencing methodologies at a cost much less than that for one year's worth of SETI. (Click here for the subsequent molecular-biological confirmation that the 20 or so cases of apparent independent convergent evolution of eyes are, almost certainly, cases of divergent evolution from one common ancestor with eyes!)
All things considered, my present guess is that fi is some million times smaller than Drake's estimate.
My somewhat facetious, necessarily uncertain, but perhaps more "reasonable" guesses of the magnitudes of fl and fi reduce the likelihood of detecting another civilization a trillion times. Morrison suggests that the SETI project has a "fair chance" of success (in Ferrisl). But right now, investing in SETI looks more like placing a bet on a dead horse. Surely, unless and until careful studies of the comparative biochemistry of "convergent" camera eyes show that fi may approach 1, no significant social expenditure on SETI is warranted.
However, in no way should this argument be construed as a criticism of NASA's rational Planetary Program which is also vigorously promoted by Sagan and Morrison.
l. F. Crick. Life Itself: Its Origin and Nature, Simon & Schuster, New York (1981).
2. L. Ornstein, Science 144, 614 (1964).
3. T. Ferris. The New York Times Magazine, 23 October 1977, page 30.
4. C. Sagan and F. Drake, Sci. Am. 232, May 1975, page 80.
5. S. Duke Elder, in Systems of Opthalmology, Vol. 1, Part 2, Mosby, St. Louis (1958).
6. L. V. Salvini-Plawer, and E. Mayr, in Evolutionary Biology, Vol . 10, M. K. Hecht, W: C. Steere, B. Wallace, eds. Plenum. New York (1977), page 207.
7. M. F. Land, Sci. Am. 239, June (1978). page 126.
8. S. Pearse and V. B. Pearse, Science 199. 458 (1978).
Return to Len's Home Page (Contingent Links)