“Of all the questions that may arise about eye movements, probably the least likely one is the simple question: Why?”

If I were to pass this question, from Walls' (1962) classic essay on the evolution of eye movements, on to you, what would you reply? Chances are that you would say that we make eye movements to look at things. That we use them to scan our surroundings. After all, if our eyes were fixed, it would be considerably more difficult to switch gaze from one object to the next. Head movements could do the job, of course, but not as effortlessly. But this post is about the evolutionary origins of eye movements, rather than the specifics of human vision. The relevant question is, therefore, whether other animals scan the environment in the same way that we do. And, to give a preview of what will follow, some do and some don't. Almost all sighted species make eye movements of some kind (or analogous head movements), but comparatively few actively look at objects the way that we do. They use their eyes to see, of course, and they may make an eye movement if something is completely outside of their field of view, but beyond this their direction of gaze says very little about what they are attending to. Assuming, of course, that they attend to specific objects at all. Therefore, eye movements must serve a function beyond scanning the environment.

The first eye movements

So what might this function be? An important clue is that photoreceptors (the light sensors in the eyes) are slow. In order for a typical photoreceptor to give a decent response it will need to be stimulated continuously for something like 20 ms (Land & Nilsson, 2002). You can think of a photoreceptor as a pixel. In a high-resolution eye, each photoreceptor “sees” only a very small part of the visual field, just like in a high-resolution photo each pixel represents only a tiny part of the image. So you can imagine what happens if an animal moves: The image of the scene will move rapidly across the retina and each photoreceptor will “see” individual parts of the scene for less than the required 20 ms. The result is degraded vision, which is a shame, because evolution has invested a lot of effort in the quality of vision. The solution is obvious. Eyes should be kept as still as possible, even during movements of the body. And this is exactly what most animals do. They have adopted a strategy not unlike that of a dancer holding his head still while making a pirouette, followed by a rapid turn of the head to catch up with the rest of his body.

Although performing a perfect pirouette is undoubtedly very difficult, and certainly beyond my own capabilities, our eyes (and to a certain extent our head as well) automatically do much the same thing. The oldest reflex to stabilize gaze is probably the vestibulo-ocular reflex (VOR). This reflex is the result of a fairly direct connection between the vestibular system, which senses rotation and acceleration, and the eye movement system. Every rotation of the body is automatically compensated by an eye movement in the opposite direction. And as a result the image of the world remains still on the retina. Like the head of a ballerina, the eyes eventually need to catch up with the rest of the body, so every period of smooth compensation is followed by a a rapid eye movement, which sets our eyes back into a more comfortable position.

The VOR is not the most precise of reflexes, which is problematic for animals with high-resolution eyes that require a very precise gaze stabilization. Therefore, there is a second reflex, called the optokinetic response (OKR). This reflex refers to the fact that our eyes are “glued” to the environment, and get dragged along when the environment moves in front of our eyes. Normally, the OKR is hard to distinguish from the VOR, since both work in concert to accomplish the same thing: visual stability. But it is easy to elicit an artificial OKR by placing an animal in a cylinder with stripes on the inside. If the cylinder rotates, the world will seem to move (even if the vestibular system, which does not fall for visual trickery, says otherwise). The eyes of the animal will follow the moving stripes, until they can't anymore, and then they jump back. This trick works so well, that it is even used to determine the resolution of an animal's eyes. By making the pattern of stripes progressively finer, the resolution of the eyes can be deduced from the point where an OKR is no longer elicited.

Scanning the environment with saccadic eye movements

So eye movements started out as a means of gaze stabilization. But as a moment's introspection will tell you, we use eye movements for more than that. The evolution of eye movements is intertwined with that of the retina (the layer in the back of the eyes that contains the photoreceptors). Evolution is always striving to improve the quality of vision. What constitutes quality depends, as I've described in a previous post, on the environment, but an increased resolution never hurts. Unfortunately, everything comes at a cost. Good eyes are bulky and, perhaps even worse, they require a huge brain to process all the incoming information.

But eyes have become mobile, pressured by the need for gaze stabilization. And this gives rise to an interesting new possibility. Rather than increasing the resolution of the entire eye, resolution can be increased in only a small part of the retina. This part is usually called the fovea, although in other species the terms acute zone and visual streak are sometimes used. This means, of course, that only a small part of the visual field benefits from the newfound high-resolution. This part is about as big as a thumb at arm's length. But fortunately the fovea can be moved around to sequentially sample different parts of the visual field, as illustrated in the video below. So, rather than taking a single high-resolution picture, we take a succession of small snapshots and somehow integrate these into a single percept. (How this works is an active area of research in itself and also the focus of my own research; Mathôt & Theeuwes, 2011)

As you can see in the video, these eye movements, which are called saccades, are fast and ballistic, as if the fast phase of the gaze-stabilizing reflexes has been adapted as a mechanism for scanning the visual field. It is not hard to see why it is important for these eye movements to be fast. For a variety of reasons, such as the slow response of photoreceptors outlined above, it is difficult to see properly during an eye movement of any reasonable velocity. So it is best to minimize “down-time” by moving very fast. And still, despite the amazing speed with which we move our eyes, we spend about an 1.5 hour each day essentially blind.

Tracking objects with smooth pursuit movements

So, if we want to place eye movements on an evolutionary timeline, it doesn't seems too far out to assume that the vestibulo-ocular reflex (VOR) arose first, followed by the optokinetic response (OKR) and, much more recently, saccadic eye movements. But there is another type of eye movement, apparently even more recent, and that's the smooth pursuit movement. Smooth pursuit is what you do when you track a slowly moving object, such as a bird at some distance in the sky. On the surface, smooth pursuit and the OKR are very similar, since they both involve smooth tracking of something in the environment. But the function of these two types of eye movements is diametrically opposite. The OKR takes the entire visual field as a reference, thus stabilizing gaze. In contrast, smooth pursuit takes a small object as a reference, thus keeping the object in foveal vision, while destabilizing gaze. Indeed, smooth pursuit requires both the VOR and the OKR to be deactivated, because they will just get in the way. This is a challenging task, which few animals have mastered completely. A nice example of an animal that exhibits smooth pursuit to some extent, but where the VOR and OKR still cause interference, is the praying mantis (Rossel, 1980). And, as you can see in the video below, even such good trackers as ourselves fall back to making saccades if an object moves too quickly.

Vergence eye movements

All animals that make eye movements are bilaterally symmetric, which simply means that the left and the right side are identical (although you could argue that box jellyfish, which are radially, rather than bilaterally symmetric, make eye movements in their own passive way; Garm, Oskarsson, & Nilsson, in press). If the eyes are not fixed on the head, the left and the right eye can, in theory, move independently from each other. Nevertheless, they usually don't. The reason for this is that, as we've seen, gaze stabilization is an important function of eye movements, and this obviously requires conjunctive (paired) movements of both eyes. Saccades and smooth pursuit do not impose this requirement and, indeed, there are animals that make disjunctive (independent) saccades. The most salient example of this is the chameleon, with its two funny turret-like eyes that move completely independently of each other.

However, many animals, including humans, do not make any large disjunctive eye movements. This is because we estimate depth by combining the images from the left and the right eye, which is called stereopsis. To do this effectively, both eyes need to look pretty much in the same direction at all times. Or to be more precise, both eyes need to fixate on the same object. This means that we are always a little cross-eyed, particularly when fixating a nearby object. The movements that control the degree of our cross-eyed-ness are called vergence movements. Since vergence movements consist of both eyes moving in opposite directions, they are disjunctive movements. But they are most effective if all other eye movements are conjunctive. In other words, the eyes should move in concert, except during vergence.

So where do vergence movements stand in the evolutionary scheme of things? There is very little to be found on this subject, but we are free to speculate. You might think that all species that have stereopsis also exhibit vergence eye movements (because the reverse is likely to be true), but this is not the case. For example, a praying mantis uses binocular cues to estimate depth (Rossel, 1983), but, with its eyes fixed on its head, it cannot possibly make vergence movements (it makes saccade-like head movements rather than eye movements). Other invertebrates, such as spiders, do have mobile eyes, so it is quite possible that they make vergence movements. But as far as I can tell, no-one has looked into this.

And what about birds? Eye movements in birds are not as strictly conjugate as in mammals. If one eye makes a saccade, the other eye will typically also make a saccade in the same direction, but it may be much smaller (Wallman & Pettigrew, 1985). In other words, if one eye takes the lead, the other eye will follow, but the coupling is not very strict. As I've already mentioned, vergence movements would seem to be most useful if all other eye movements are strictly coupled. Therefore, it is doubtful whether vergence movements are as common in birds as they are in mammals. Nevertheless, some birds do exhibit vergence movements (Martinoya, Le Houezec, & Bloch, 1984).

Humans, of course, show vergence movements as well. As a matter of fact, all mammals probably do, even laterally eyed animals such as rabbits (Zuidam & Collewijn, 1979). But the best clue to the evolutionary place of vergence movements is that all animals that show vergence movements show the full range of other eye movements as well, whereas the reverse is not true (think of the chameleon). This strongly suggests that vergence movements are new, quite possibly the most recent addition to the family of eye movements.

Convergent evolution

There is surprisingly little variation in the type of eye movements that is found across the animal kingdom. So far, we have seen the gaze-stabilizing reflexes (the VOR and the OKR), saccades, smooth pursuit and vergence. All of these occur in a wide range of species, although the “new” eye movements are less common than the ancient gaze-stabilizing reflexes. A praying mantis makes saccades and shows tracking behavior, much like we do (Rossel, 1980). This is remarkable, considering that our most recent common ancestor (or “concestor”, as Richard Dawkins likes to call it) lived about half a billion years ago. And this concestor did not have a sophisticated eye movement system that we and the praying mantis could have inherited, thus explaining the similarity. So the story of the evolution of eye movements is one of convergent evolution. Over and over and over again, nature arrived at the same solutions.

But, even though the repertoire of available eye movements is limited, not all animals possess all eye movements. We like to think of ourselves as the pinnacle of evolution, but there are at least two types of eye movements that we do not possess.

Like a vergence: “field switching”

Prey animals, such as rabbits, typically have lateral eyes, placed on the side of the head. This gives them the obvious advantage of having an almost panoramic field of view (Hughes, 1972), which makes it pretty much impossible for a predator to approach a rabbit undetected. But this comes at a cost in the form of reduced depth perception. To maximize the full field of view, there must be little overlap between both eyes. Consequently, there is only a small part of the visual field where stereopsis can be used. So, if you're not being preyed upon, it's better to have frontal eyes, like us, with a large binocular overlap. If only you could switch from lateral to frontal vision at will...

And some animals can (Wallman & Pettigrew, 1985)! The thawny frogmouth, a small bird that is related to the owl, has eyes that are more or less lateral. This makes sense, because it is being preyed upon by other, bigger birds. But it is a predator itself as well, so it could use a good pair of frontal eyes for hunting.

Amazingly, the thawny frogmouth, and presumably quite a few other birds as well, switches between frontal and lateral vision by making large, vergence-like saccades. Normally, it will be in “panoramic mode”, on the lookout for predators. But if it spots a potential treat, such as a mouse, it will go into “frontal mode” and start to look very much like an owl (you can see the switch quite nicely at around 0:24 in the video).

These “field switching” vergence eye movements are similar to the more familiar vergence movements, described above, in the sense that both eyes move simultaneously and in opposite directions. But they are very different in the sense that they are large and saccade-like. And presumably they have little to do with depth perception (although conventional vergence movements may occur when the bird is in frontal mode).

Line-scanning with a narrow retina

The last type of eye movement that I'll discuss is perhaps the most remarkable, because it is so remote from our own experience (mostly taken from Land & Nilsson, 2002). Jumping spiders are famous for their complex and flexible behavior. They are visual hunters that scan the environment for prey. When they have located a potential meal, they will creep up on it, preferably from behind, and subdue it with a venomous bite. Jumping spiders have four pairs of eyes (so that's eight in total). One pair in particular is used to identify prey. The remaining eyes serve mostly as motion detectors. This is, incidentally, a variation on a familiar theme: We use peripheral vision to detect motion and switch to foveal vision when we have detected something of potential interest. The strategy of jumping spiders is much the same, but rather than switching between the periphery and fovea of the same eye, they switch between eyes.

But the primary eyes of jumping spiders have another remarkable characteristic. The retinas of these eyes are boomerang-like structures, with very narrow, elongated fields of view (Land, 1969). These narrow retinas have given rise to a rare (although not unique) type of eye movements: slow scanning movements. In an eye movement that superficially resembles smooth pursuit, jumping spiders let their eyes slide across whatever object they are observing. But these are not tracking movements, since the object under observation may be perfectly still. The spider simply scans the object, much like a photocopier, one line at a time.

(In the video above, you can see the spider moving its eyes. It actually moves only its retina, the chamber is fixed on the body. I'm not sure that this specimen is of the "line-scanning" kind, but it is a nice video anyhow.)

Recap

Vision and mobility don't go well together. If an animal moves its body, the retinal image of the world changes dramatically. And the quality of vision suffers as a result. To alleviate this problem, eyes move to compensate for the movement of the rest of the body. For example, when you shake “no”, your eyes will automatically cancel out the rotation of your head and you will have no problem in keeping your eyes fixated on the person in front of you. The first eye movements are likely to have been these type of gaze-stabilizing reflexes (Land, 1999; Land & Nilsson, 2002; Walls, 1962). Therefore, in a sense, gaze stabilization is the primary reason why we move our eyes.

Of course, eye movements are no longer used solely for gaze stabilization. Animals scan their environments using saccades, track objects using smooth pursuit, estimate depth using vergence movements, and so forth. But the real surprise is that there is not more variation. The way that a praying mantis scans its environment, for example, is uncannily human-like. And this is not cherry-picking. The eye movement repertoires of distantly related species often show remarkable similarities. The evolution of eye movements is really a story of convergent evolution.

References