squids and fishes

Evolution has a reputation as a genius, and it certainly is skilled at designing organisms. But is it strong in every way, or does it have weaknesses? In 1997 I made an informal investigation by comparing separately-evolved systems for the same purposes in squids and fishes, and concluded that evolution is mostly strong but has a major weakness.

You may argue that I am biased. The weakness I found is premature convergence, which is well-known in artificial intelligence as the great weakness of artificial evolutionary algorithms. Perhaps I only found what I was looking for. I certainly did not nail down details anywhere near tightly enough to call my result a scientific conclusion. But I still find it a convincing informal argument. And though I did not notice many scientists agreeing with me in the 1990’s (whether that’s their fault or mine), in 2013 I see biologists increasingly emphasizing evolution’s limitations.

Some of the science has changed since I researched this in 1997. See the last section for a 2013 science update and a new argument.

I have been reading about cephalopods, the class of mollusks that includes squids, cuttlefishes, octopuses, the chambered nautilus, and extinct animals like the belemnites and the ammonites with their fabulous coiled shells.

I want to know all about squids and octopuses because they’re the most alien intelligences on earth: They are the only invertebrates with active life styles and big brains like vertebrates. It may seem weird to be inside a bat’s head or a dolphin’s head, but compared to a squid those two are your cousins. And intelligence is what I most want to understand—it’s the enabling technology of all other technologies—so I want to know all its different forms. The workings of neurons were first figured out in the squid giant axon, but even so not enough is known to get into a squid’s head. So rather than finding out about the second form of natural intelligence, the brain, in practice I’ve mostly been finding out about the first form, evolution.

Evolution is the foremost believer in Orthotropous Max: It drives everything to the limit, even if it has to turn scales into feathers. Evolution has a reputation for creativity and thoroughness; organisms that evolved are thought to be superbly designed. Here’s a typical quote: “The effectiveness of evolution is so well established that where the design of a fish seems bad it is probably only because we have failed to understand it.” — R. McN. Alexander

But is it true? Fishes and squids evolved separately, but they are direct competitors and both need high-performance systems. The pair make a good test case. If unrelated evolved systems are equally well-designed, then evolution lives up to its reputation.

The nautilus is a living fossil that’s only distantly related to the interesting squids, cuttlefishes, and octopuses. So I’ll call the squids, cuttlefishes and octopuses by their technical name, the coleoid cephalopods. I compare them to the ray-finned fishes (which is most kinds, in this evolutionary era).


The nautilus lives by smell and has primitive pinhole camera eyes. Coleoids live by vision and have sophisticated eyes about as good as fish eyes (which are rather different from the eyes of mammals, because they’re for seeing underwater). The overall designs are similar: Both have an eyeball that contains a rigid spherical lens with a graded refractive index and a retina, and accomodate by moving the lens. Virtually all the details are different. For example, the lens of the vertebrate eye is made of cells, and most fishes change focus by moving the lens horizontally, so that different parts of the visual field are simultaneously focused at different depths, a rather odd arrangement. The coleoid lens is secreted and focuses by moving in and out, like a camera lens. The vertebrate retina is inverted, with nerve axons passing through the light path in front of the rods and cones and exiting the eyeball through a blind spot. This makes sense in fishes because they swap the rods and cones in and out of the plane of focus as part of dark/light adaptation, and putting the nerve axons behind the receptors would interfere with the movement. The coleoid retina has rod-shaped receptors only, and the nerves leave through the rear, a simpler and sensible arrangement.

Many vertebrates can see in color, while apparently only one species of squid can. Cephalopods can see the plane of polarization of polarized light, which must be handy for seeing through the reflective camouflage of fishes. I ran across a report that fishes can detect polarized light too, but I could not find any details so I’m not sure it’s true.

So the two systems are pretty even in capability, but fishes may have an edge. Presumably they get some advantage from their complicated way of adapting to dark and light—I trust evolution that far.


Cephalopods have no sense of hearing. Sound is more important in the water than on land, so it’s a strange lack, especially since a major class of predators, the toothed cetaceans, search out squid to eat by sonar. Coleoids, at least, can detect low-frequency sounds with the lateral line and the statocysts, which I talk about below. Many fishes have sensitive hearing.

lateral line

Fishes have a lateral line organ which detects water movements and the lowest frequencies of sound. A fish lateral line can detect a meter-long fish thirty meters away, so it’s quite sensitive.

Coleoids have a similar system, with lines across the head and down the arms. The sensitivity of the receptor cells has been measured and found to be the same as in fishes.

vestibular system

Vertebrates measure angular accelerations and the direction of gravity with a vestibular system that includes a set of three orthogonal semicircular canals. Cephalopods do the same job using statocysts which work differently. Octopus statocysts include a low-rate turn sensor for walking and a high-rate turn sensor for swimming, but don’t bother to detect turning in all three axes (since octopuses live on the bottom). Squid statocysts sense in three axes, and have been shown to work as well as fish semicircular canals.


Many fishes can change color to a limited degree for camouflage and signalling. A few fishes, like flatfish, are quite good at changing to match the background color and texture. Fish chromatophores involve large branching cells which contain pigment granules that can be moved around by intracellular mechanisms. When the pigment is spread out in the cell’s branches it’s visible, and when it’s concentrated it’s not.

The coleoids change color better than all other animals. They have chromatophores which are worked by muscles and controlled by a dedicated lobe of the brain. They can change color instantly, display moving patterns, and display different patterns on different parts of their bodies. The repertoires of shallow water species tend to include things like sandy and rocky bottoms (they tune the pattern so they match as closely as possible), seaweed, loose algae, a harmless fish (for sneaking up on you), and a predatory fish (for trying to scare you away). They also have special displays for communicating with each other. I found mention of a seven-hour videotape of an octopus prowling for food, in which the octopus changed its colors on average once every thirty seconds.

Each chromatophore is a little sac of pigment with radiating muscles. When the muscles are relaxed, the pigment sac is small and invisible. When the muscles pull outward, the sac flattens and the pigment is displayed. It’s simple, and many animals could make good use of such a thing if they had it. You have to wonder why it has only evolved once.

digestive system

Digestive systems are hard to compare because species have different needs. Coleoids have a live-fast-die-young strategy that requires fast digestion. Some species grow as much as 10% per day averaged over the lifespan, an outrageous rate that no other cold-blooded animal can touch. Fishes are more relaxed about their growth rates and can afford to take more time to digest.

The cephalopod digestive system processes food in batches rather than continuously. Food is munged up in the stomach and passed to another pouch, called the cecum, for absorption. Loligo squids digest a meal in four to six hours, about the same as the fastest fishes and several times faster than most fishes. The fast-digesting fishes do it by warming up the gut above the water temperature, hard enough for a fish and totally out of the question for a squid. Apparently squids found another way.

I could not find absorption rates for fishes: What proportion of protein they assimilate, and all that. The rates I found for squids are quoted with respect, so I gather they’re respectable.

immune system

The vertebrate immune system includes a nonspecific response to anything it recognizes as bad, and a complicated specific response to nasties that it has seen before. It works pretty well.

After a long search, I finally discovered a 1996 reference that says that virtually nothing is known about the immune systems of cephalopods. With their short lifespans, squids would not get much use out of a specific immune system. Other mollusks have effective nonspecific immune systems. The general rule seems to be that, for any given infection, most members of a mollusk population will resist it totally and a small proportion will be susceptible. So this seems to work pretty well too, but that’s the best I can say.

circulatory system

Cephalopod blood is transparent because its oxygen-carrier is hemocyanin, based on copper. Vertebrate blood with hemoglobin is several times better at carrying oxygen. I found one source which attributes the poor performance to the large size of the hemocyanin molecule—it uses many more amino acids per oxygen binding site than hemoglobin. Another source blames it on the fact that the hemocyanin is loose in the blood rather than packed up into blood cells. That means the blood can’t carry as much of it without becoming too viscous.

This is a big deal. Squids have to pump blood around their bodies at a ferocious rate to keep up with oxygen demands, they can’t sustain work as long as fishes can, and their entire biochemistry has to be optimized for oxygen efficiency at the expense of other considerations. It’s got to be a limiting factor in what niches they can occupy and how successful they can be.

Other aspects of the circulatory system seem to be comparable in squids and... not fishes, but the hot-headed energy-exuberant mammals, which share flamboyant heart rates. Squids have three hearts arranged in a way that’s functionally similar to the mammalian four-chambered heart. Mechanical properties of the blood vessels have been shown to be similar.

nervous system

Vertebrate brains share a common ground plan. A fish brain is like a simplified, cut-down mammal brain. Cephalopod brains are organized in a ring around the mouth, like the arms. The esophagus runs through the middle of the brain. The nautilus brain is like a simplified, cut-down version of the coleoid brain. I think it’s interesting that brain plans have been conserved for such a whopping long time (coleoids and nautiloids diverged in the Paleozoic, long before dinosaurs evolved, and so did fishes and amphibians).

Coleoid brains, relative to body weight, are bigger than fish brains but not as big as mammal brains. Most likely they are smarter than fishes and not as smart as the notoriously intelligent marine mammals. Octopuses have been subjected to countless behaviorist learning batteries, so there’s some solid evidence here. It makes sense that coleoids need to be smarter than fishes, because they have more choices. They have to control all their arms, and decide what colors to be and whether to swim forward, backward or sideways.

We’re no good at measuring the complexity of behavior, so as far as we can tell a squid brain is as good as any, pound per pound. Maybe it is, but there’s one way that vertebrate brains are unquestionably better: myelinated axons. Only vertebrates evolved this trick for vastly speeding up nerve conduction. I think it is a big deal, because even people have special adaptations to speed up reactions, such as spinal cord reflexes and the half-hardwired vestibulo-ocular reflex. I take this to be evidence that even the higher nerve speeds in vertebrates are not as high as evolution would like; you can always use faster reactions.


I didn’t find enough information to compare other systems, like the gills, the sense of smell, and the efficiency of muscles. This set should be enough to draw the vague conclusions I’m interested in.

I have made only feeble efforts to double-check facts. For the facts that I have checked in more than one place, my sources disagreed amazingly often. Otherwise I quoted details from memory, and my memory is atrocious so I’m sure I messed up a bunch of them. You might well conclude that you can’t believe a word of it.

The worst is that the whole comparison enterprise is methodologically dubious. We don’t thoroughly understand any single system in either fishes or squids, and we don’t understand all the adaptive tradeoffs that go with possible variations on the systems, and we don’t understand the overall adaptive strategies of fishes and squids. What I said about the digestive system, that squids may have better digestive systems because their lifestyle demands it, could apply to all the other systems too, and we haven’t figured it out yet.

The bottom line is that it’s not possible to draw scientific conclusions. I’ll have to make do with heuristic conclusions.

the winner

On the scorecard, fishes win points for hearing, hemoglobin, and myelin, and a quarter point for possibly superior eyes. Squids score on color changing, plus half a point for possibly more efficient digestion. The two struggle to a draw on the lateral line and the vestibular system, and the immune system match is undecided. The winner and still champion, since the late Cretaceous, fishes!

Fishes are dominant nowadays. There are over 20,000 species of fish, and only 700 or so species of cephalopod [as counted circa 1997]. Admittedly most of the fish species are freshwater, since every river and lake can have its own local species, but fishes have more diversity even ignoring the freshwater ones. Fishes are tops in upper waters, a more favorable habitat because there’s more food. Squids are tops in midwater. Uniform food-sparse habitats favor having few species, each with many individuals, which is the case for midwater midocean squids.

That fishes win my comparison does not prove that they are better adapted than squids. I did not compare the most important factors.

One factor is cost of transport. Thanks to basic physics, fish swimming is several times more efficient than squid jetting at moderate swimming speeds. Jetting has countervailing advantages, tremendous flexibility and greater acceleration in emergencies, but low cost of transport would seem to be competitively devastating: Fishes do not need to eat as much as squids do, they can undertake long migrations more easily, and so on. Cost of transport is the factor I see cited as the reason that squids have fast growth and short lifespans; it constrains their adaptive strategy.

Cost of transport is the most important factor I can think of behind the Fish Ascendancy, but I have two reasons to believe that more important factors exist, even though I don’t know what they are.

1. Illex illecebrosus squids live in the open ocean and make their living swimming at high speed between shoals of fish which they catch. I don’t know how this is possible, because for this lifestyle cheap swimming would seem to be essential. Fast-swimming fishes should eat their breakfast. I can only speculate about how the squids make up for their disadvantage. Maybe they’re better at catching fishes than fishes are.

2. Fishes and cephalopods both evolved in the Paleozoic, early in the history of animal life, and they’ve probably been competitors ever since. It’s hard to be sure of anything when fossils are your only evidence, but over most of this colossal interval, it looks to me like cephalopods were more successful than fishes. The teleost fishes underwent an adaptive radiation in the late Cretaceous, and that’s the earliest you could say that they established dominance. Most of dinosaur time was owned by belemnites (which were squid-like and either are or are closely related to coleoids) and ammonoids (which are nautilus-like but are considered to be more closely related to coleoids than to the modern nautilus), both extinct.

So apparently coleoid cephalopods (or a slightly wider evolutionary group that includes them) are more successful than earlier fishes but not as successful as modern teleost fishes. I do not know what evolutionary breakthrough or environmental change gave the teleost fishes their chance; I’ve never looked into it.


Going back to the scorecard, there are nine points under contention. The fishes win cleanly on three counts, the squids on one count, and five are even or uncertain.

The systems that come out even, or nearly so, are evidence of evolution’s strength. The independently-evolved vision, vestibular, and lateral line systems are complex, but they come out with equally good designs with equal performance, a fantastic beautiful result that must be due to successful orthotropization, not to chance. All kneel! Evolution is the preeminent designer.

The systems that come out with a winner may be evidence of evolution’s weakness. When evolution has an idea, it can extend and fine-tune like nobody’s business, but it does not have all the ideas. For squids it did not have the idea of a sense of hearing, or of hemoglobin, or of myelinated axons. For all non-coleoids it did not have the idea of muscle-actuated chromatophores. Evolution is creative and has a lot of ideas, but by no means all of them, or all of the important ones.

There are methods in artificial intelligence, such as genetic algorithms, which are inspired by natural evolution. The big problem with evolutionary methods in AI is called premature convergence. It happens when your artificial population settles on a solution which isn’t the best one—and having found a solution stops exploring, so that it can never find a better one. If squids were the outcome of an AI experiment, I’d say that hemocyanin was the result of premature convergence on an oxygen-carrier without adequate exploration of the alternatives. Real evolution and artificial evolution are unlike in many ways, but I conclude that they share the same principal limitation.

The good ideas that evolution hasn’t had yet can be called constraints: They constrain an organism’s design and its ability to adapt and compete. When evolution has the new idea, the constraint is broken and new evolutionary possibilities arise. Some constraints may be so firmly set that they can’t be broken. For example, no animal could evolve a radically novel way of encoding its DNA, because too many other parts of the animal’s workings depend on the current way. (Perhaps a micro-organism could, though even that seems tricky.)

There once lived an animal called Urbilateria which was the common ancestor of all bilaterally-symmetrical animals (this excludes sponges, jellyfish, and some others). It may have been a worm. This first animal is known to have had muscles, a gut, a nervous system, and eyes (probably eyespots). Every muscle, gut, nervous system, and eye that has evolved since then is derived from the first animal. If Urbilateria prematurely converged on a bad solution to some problem, and this emplaced a constraint that could not later be broken, then it might be that all animals are badly designed. What an alarming idea!

I think this happened in the case of nervous systems. Nervous conduction works in an awkward and inefficient way, which might have been fine for Urbilateria but is clumsy for larger, smarter animals. It is orders of magnitude slower than electrical conduction. I suspect that evolution could have done better if only it had had a luckier idea in the first place. Vertebrates have the fastest signal conduction, but even so they have adaptations that look to me like design compromises to speed up their reactions as much as possible.

There are other good ideas that evolution has never had. For example, radio. Radio sensing has probably never evolved not because it’s hard to implement (think how simple a crystal radio is), but because radio doesn’t tell you much about your environment. By comparison, radio sensing plus radio transmission would give an animal not only a communication channel but also radar, which seems invaluable. Now that humans have started putting transmitters all over the place, maybe there is adaptive value in radio sensing and it will start to evolve—let’s wait ten million years and find out!

I can’t help wondering whether basic mechanisms of life, like DNA encoding and protein transcription, are as good as they seem to be. Are ribosomes as well-designed as we think, or are they finely-tuned but busted at heart, like nerve conduction? Were they perhaps well-designed when they arose billions of years ago, but poorly suited to the modern world? The adaptive tradeoffs behind these systems are still well beyond our ability to understand, so we can’t double-check them yet.

The bottom line is that evolution is extremely impressive but far from all-capable. To create human beings starting with nothing more than combinatorial chemistry is mind-boggling. Evolution is so amazingly skilled that I think we tend to overlook its big oversights.

I had fun learning all this stuff and thinking about it. I rather imagine that my cavalier treatment would appall an evolutionary biologist, but maybe I’ll find out about that someday.

January 1998 reply to an objection

One person complained “I’m a little perturbed by your basic attitude that evolution has any particular purpose in mind.”

There are lots of weaknesses in my argument, but this isn’t one of them. I don’t ask for evolution to have a purpose, only an effect.

I quoted R. McN. Alexander as writing, “The effectiveness of evolution is so well established that where the design of a fish seems bad it is probably only because we have failed to understand it.” He thinks that an animal could be well-made or poorly-made. In other words, there’s a standard that we can hold evolution to: Does it make good or poor animals? And he concludes that it makes good animals. Nothing controversial here. All I’m doing is bringing different evidence to bear on the same question.

I can say the same thing more precisely, with a dose of idealization. Imagine the mathematical space of all possible animals, which includes star-nosed moles, quivering cubes of flesh that will die in a few minutes, weird tube worms that live under the ice of Europa—all of ’em. At a given time, the population of any real species in some habitat will occupy some teeny nook in this space.

Now wait a generation or so and look at the population again. It will occupy a slightly different nook, even if the environment was constant, partly because of genetic drift and partly because of natural selection. Now imagine that you can repeat the experiment by replacing the population in the same habitat and same nook and running time forward again with slightly different results. In this way, in principle you can disentangle the effects of genetic drift and natural selection. It may be a little harder in practice, unless you have a time machine.

Since in this thought experiment you can gauge the effects of natural selection, you can define “better” and “worse” regions of the space of animals, for a given habitat. Put a population in a worse region and it’ll be subject to selection pressure moving it toward a better region (or it’ll die out). The better and worse regions are decided by a whole mess of factors. The population’s variation determines what natural selection has to act on. The physical environment of the habitat, the other organisms there (food, competitors, predators, parasites, dangerous critters to be avoided, harmless passers by), and the animal population under consideration itself and all its interactions with the others, whew, determine the actual selection that is done. Then the heritability of traits and the genetics of the animal determine the next population, and so on. And this is clearly idealized, because in practice the population itself has effects on the other animals and the physical environment, so the adaptive landscape of the next generation is different from the current one’s. But it’s close enough for SWAPA work.

Now I can try to define “good design” and “bad design” in Alexander’s sense. The context I’ve put it in is so precise that this idea is fuzzy in comparison. But you can imagine that you could try different animal designs in the same habitat, with everything else held constant, and somehow measure which design is “more successful” in terms of population biomass, or a measure of resource exploitation, or ability to overcome a competitor species, or whatever you like. If you can do this, then you’ve defined a “goodness” function over the entire space of animals (for a specific habitat), and you’ve made Alexander’s statement meaningful. You might be able to say that evil American gray squirrels are better than European brown squirrels in a specific sense, for example.

Variation, selection, and inheritance work on the scale of a few generations to move the population toward a nearby “better” region, if there is one in the range of variation. That’s evolution. The question I addressed is this: What is the absolute level of goodness that evolution, including genetic drift, over geologic time and with all the habitat changes that have occurred over geologic time, has brought organisms to? In other words, how effective is the local optimization of evolution at eventually moving animals to a global optimum, or at least to an excellent regional optimum?

And I did that by choosing two unrelated kinds of animal that live in sort of similar niches, and comparing complicated subsystems that I sort of understand, like their eyes and blood pigments, to see if they seem to come up to sort of the same level of goodness, which I didn’t actually define but made up as I went along. If they do, and there’s a close match as in the eyes, that’s suggestive evidence that that system is near a regional optimum. At least it suggests that there’s a large basin of attraction. It’s remarkable to see independently-evolved sophisticated eyes with such similar designs. If there’s a big mismatch of goodness, as in the blood pigments, that’s evidence that one of the animals is far from a global optimum.

I chose this idealized theoretical level of abstraction to make my argument’s workings utterly clear. The “purpose” that you saw has vanished. I hope I’ve straightened out my viewpoint.

2013 update

new science

When I researched this in 1997, all my sources agreed that the hemocyanin blood pigment of cephalopods was inferior to the hemoglobin of fishes. They gave measurements to back it up: For example, because hemocyanin carries oxygen at lower density, cephalopods have to pump a much higher volume of blood to oxygenate their tissues, six times higher in one measurement. That’s a clear disadvantage.

Today it is known that hemocyanin has a countervailing advantage as well. When oxygen levels are low, hemocyanin carries oxygen more efficiently than hemoglobin, with the result that cephalopods can remain active at low oxygen levels where fishes are at a disadvantage. This shows both in their ecology today and in their evolutionary history. Fishes live at all depths, but dominate the surface waters where oxygen is high. Squids live at all depths, but are predominant at greater depths where there is less oxygen.

Cephalopods and fishes both first evolved in the Cambrian at the beginning of the Paleozoic, and must have already chosen their blood pigments. Back then, oxygen levels in the ocean were much lower than today, so fishes chose “wrong”. Sure enough, cephalopods diversified in the following Ordovician period to become the largest animals and top predators, while fishes continued to evolve but remained lesser actors. But much later, in the Devonian, vascular plants appeared on land and oxygen levels rose. Fishes were now “right” and diversified spectacularly, so that the Devonian is called the Age of Fishes. At the end of the Paleozoic a devastating mass extinction caused oxygen levels to crash again, and cephalopods (especially ammonites) recovered faster from the extinction than fishes. Today oxygen levels are high again, and fishes dominate the best habitats.

This change in understanding doesn’t undermine but reinforces my original argument. Each lineage chose its blood pigment early and has been unable to evolve to change it for over 500,000,000 years even when there was a compelling need. Sometimes by luck one line benefits from changes to the environment, sometimes the other, and both are at the mercy of outside events because evolution has locked in a key variable.

I shouldn’t overstate it. There are examples known of animals that have changed out their respiratory pigments, but it does appear to be rare. The change is not impossible.

Oh, and another advantage of hemocyanin: It doesn't have any strong color. Some squids make use of that fact to be largely transparent, which must help their camouflage.

another argument

Seagrasses are flowering plants that live in the ocean. The evolutionary trajectory is something like this: Green algae lives in the ocean. It adapts to freshwater, then eventually colonizes land and evolves into a a plant similar to moss which reproduces by airborne spores, then later gains height and eventually develops xylem and phloem to transport water and nutrients, becoming a vascular plant. Later come seeds and after a tough evolutionary slog, over a hundred million years after seeds, flowering plants show up. Whew. Finally, perhaps toward the end of the Cretaceous, the earliest seagrasses shift from living in freshwater to living in the ocean, perhaps moving down the rivers in a reversal of their origination hundreds of millions of years earlier.

A fabulous evolutionary success. What does it tell us about evolution’s failures? Well, a species can only evolve into a new niche if either it has an advantage over the current inhabitants—or if the niche is empty. I believe that seagrasses evolved to occupy an empty niche that no marine algae already occupied. Seagrasses compete with microscopic algae, but large algae doesn’t grow where seagrasses do.

Multicellular red algae has existed for at least 1,200 million years. Green algae fossils are known from the Cambrian, over 500 million years ago. So, with red and green algae having had opportunity over geological time, how could the seagrass niche remain empty? Or if the niche wasn’t empty, then how did a land plant outcompete some algae that was on its home ground and should have been ideally adapted? Evolution has the weakness that it can only operate in small steps. Life walks to the edge of its fitness limit, but it can’t look beyond. There is a long sequence of short steps that runs from green algae to seagrasses—we know because seagrasses took them—but there may not exist any such sequence that stays entirely in the ocean. You can’t get from there to here without taking a detour to collect different adaptations.

The argument is not watertight. Maybe seagrasses took advantage of an environmental change. But on the face of it, it’s an example of the same weakness of premature convergence that shows in the case of squids and fishes. A successful solution may be discovered early and never change much, like red algae which still looks the same as it did in Precambrian time even though the competitive environment and the ocean’s temperature and chemistry have never stopped changing.

originally written November 1997 and January 1998
added here January 2012, last update September 2013