Octopus (23 page)

This kind of field observation can be used for constructing an interesting lab test of spatial learning. Jean Boal et al. (2000) kept octopuses in a tank that had two small deeper areas at one end, and gave them a few days to explore their new home. Then they drained the water out of the shallow part, leaving two deep refuges, like tide pools. The octopuses remembered the locations of the deeper areas and learned to go to one of them until the water was restored to the rest of the tank, as would happen when the tide comes in (octopuses can survive for a while in air but prefer to stay in water). It's a good memory test, using an ability the octopus has to practice each day.

Octopuses don't hunt in areas they have covered in the past few days. This win-switch foraging strategy is very efficient. If an octopus catches a crab under a particular rock, it shouldn't bother going back to that spot because there won't be another crab there for a while. This behavior in the octopus supports the idea that it has working memory; it remembers not just where things are, like its home, but it also remembers where it has gone for food. This situation has been used in testing animal memory. Rats were given a radial maze that has several arms built out from a central platform, each with food at the end. The rats could go anywhere to get food, but they learned not to revisit arms already plundered.

We made a water version of a radial maze and tried to prove that octopuses had this ability in the lab too, but the octopuses didn't seem to learn to go to still-baited arms that they had not explored before. Later I learned why this experiment didn't work. I tested the small and easily available red octopus, but it turns out that this species may not return to the same home day after day. The test might have worked with a common octopus or a Hawaiian day octopus, both of which do.

This win-switch pattern of foraging areas gives us a clue as to why a lot of learning researchers might have had trouble getting octopuses to make consistent choices of the rewarded rather than the unrewarded stimulus. The octopuses would only choose correctly seven of ten times on a series of trials. Rats made more and more positive choices the longer scientists
tested them, and eventually got to a perfect ten out of ten. Octopuses just kept randomly trying the other, unrewarded, choice once in a while. This makes sense from the viewpoint of an octopus that has lived in the sea for months before it was brought into the lab. Food isn't available again in the sea where you just found it, so why try? In a week or so, it might work, since prey animals always need shelter. So octopuses keep trying the alternative choice some of the time, because that approach works out in the ocean.

Octopuses also use learning to direct their actions. A simple type of learning is habituation, when an animal stops responding to some repeated and uninteresting situation. Humans use habituation all the time—no longer noticing the hum of the air conditioner or hearing the piano practicing coming from next door unless it's particularly off key. Octopuses may well use it too. Many animals, including the simplest worms and sea anemones, will habituate to repeated stimuli in their environment. But we found that habituation had been little studied in octopuses, so we, and Michi Kuba and colleagues (2006), took it on.

Kuba continued to study learning in common octopuses along the same lines that we had studied earlier in giant Pacific octopuses, showing their balance among exploration, or getting information out of the environment; habituation, or losing interest; and play, or manipulating the environment or the social interface without an obvious immediate purpose. It's seems logical that octopuses would explore, since they have to get information to survive in their variable environment. In the studies, the octopuses started off a sequence of trials by grabbing a floating pill bottle or three-dimensional plastic block we gave them. Then they pulled it to the mouth and held it with all the suckers on the arm bases. As you would expect, after a while the octopus held it less tightly, eventually lost interest, and stopped exploring it as the trial wore on.

When we gave the octopus a test object such as a jar (see plate 31), it would explore it until it got tired of the object. When Kuba used as his stimulus object a plastic crab model outside the tank, the octopus quickly learned it had seen it before and couldn't do anything with it, and lost interest. But when the octopus could manipulate the object, results for habituation among tests wasn't consistent—there was octopus variability. Part of the reason that habituation wasn't consistent is that we began to see play, another action only higher vertebrates are supposed to do. Animals with a lot of manipulative ability are able to move in viewpoint from
“What does this object do?” to “What can I do with this object?” Instead of losing interest, they switch approach. The octopus Roland first saw doing this play behavior blew jets of water from her funnel to send the pill bottle to the other end of the aquarium where the water intake flow sent it back to her, and she repeated this action twenty times. He didn't see a consistent decline in response, but instead saw a change in the type of response. The playlike behavior peaked at the third or fourth day, then the octopus seemed played out and decreased the contact time.

Can we say that an octopus is intelligent by seeing it displaying playlike behavior? Playing with objects isn't about learning, in a simple sense. An octopus will learn about an object or its own ability when it plays, but it also will do so when it just explores an object. So far, only animals that play are ones we class as intelligent, though this may be another case of anthropocentric humans needing to look further. Still, this whole set of behaviors makes us see that the octopus is intelligent. If we want to understand how animals gather and use information, we have to look at variable exploratory behavior. An octopus's going out to see what's there, feeling around to understand what things are made of, then manipulating an object to see what it can do with it is interesting behavior for study.

Another important question about intelligence and the role of learning in animals' lives is how an animal changes across time and with development. This question has barely been looked at in octopuses. But cuttlefish have turned out to be a good subject, because the hatchlings are fairly large for a cephalopod and settle onto the bottom quite soon after birth. They are unlike octopus paralarvae, which are usually tiny and float off into the plankton for weeks, and even octopus benthic juveniles are very small.

It is easier to evaluate prey capture in newly hatched cuttlefish, which are very picky about their food. John Messenger showed in 1977 that they would only take tiny mysid crustaceans or models that looked like them. When mysids were put in a tiny glass test tube and the little cuttlefish struck at them with their tentacles, they couldn't learn to stop. At the beginning of life, these animals had a narrow, fixed search image. They did not need to learn what food was appropriate to catch but they couldn't learn about new food either. As several weeks passed, the cuttlefish began to learn to try for different prey species—their search image expanded. They also learned better and better not to try to catch the mysids in the test tube. The cuttlefish were able to accomplish short-term learning, over
about five minutes, after one week of life, and they got the ability to learn long term, over one hour, by one month of age.

We believe that learning is an ability that's found in the brain, and visual learning in cephalopods is related to the brain's vertical lobe. If the lobe is surgically removed, an octopus can no longer learn. Messenger also discovered that newly hatched cuttlefish had only a tiny part of their vertical lobe. When several weeks passed and the cuttlefish could learn, anatomy studies showed a big development of the vertical lobe both in size and connections to other brain areas. Gradually a learning-dependant animal learns to eat more variety of what's around it as it has time to find it. As well, both an enriched environment and the presence of particular prey species would change the learning program in those tiny animals, so later they accepted those prey species more readily. Environment had a tuning role for learning what to eat, right from the start.

Recent research has shown some early learning in cuttlefish. Ludovic Dickel et al. in 2006 found that cuttlefish aren't always tightly bound to a narrow recognition of mysids as prey. In their first week of life, tiny cuttlefish still in their translucent egg capsule could see general shapes. If they were shown small crabs, they later changed their prey choice and also accepted crabs. Such early learning is much like that of humans, and could be called imprinting.

This learning behavior has a clear parallel with that in young mammals, including humans. At birth, humans have a set of reflexes—automatic actions that don't get changed by learning and help us to survive. We can grasp with surprising strength, which helped our monkey ancestors not to drop from mother's chest to the forest floor. We also have a sucking reflex that helps us benefit from the easily available and nutritionally perfect milk of mother's breast. In fact, newborn human (and many other mammal) babies can do something that adults lose the ability to do: suck liquid and breathe at the same time. They roll up their tongue into a flexible tube near the roof of the mouth to send milk down one way and receive air down another. These reflexes and more are gradually lost by six months of age as the baby also begins to learn to tune feeding and response to caregivers and to what is around them. Actually, they can even tell their own mother's milk from anyone else's by one week of age (Charles Maurer and Daphne Maurer 1988). The beginning stage with automatic responses and no learning is followed by a gradual replacement with learned associations. The shift is the same one experienced by cuttlefish, the best developmental
program for survival of an inexperienced newborn that will later become a learning specialist.

This study is an example of a larger area of research known as constraints on learning. Scientists have found that, generally, animals get information from their environment but that heredity controls what and when they learn. Small birds learn who their mother is, and human babies learn the visual cliff—something that looks far away, and down, and shouldn't be walked to—and that each animal has specific times of development. We can imagine that octopuses have these time-sensitive periods, such as when they settle out from the plankton. Settlement is likely a sensitive period of fast learning for young octopuses. Settling paralarvae of small-egged octopuses would have to learn where to hide, how to color match the background (which in tiny cuttlefish is affected by experience), what kind of food is edible, and even how to walk and catch prey with their newly elongated arms.

Regarding the program of learning, an octopus or any animal that depends heavily on learning will have its learning guided by heredity. After all, learn wrong and you may be dead. We have gradually learned about these constraints on learning. Bees learn about flowers' color, smell, and shape so they can find them again, but they learn some colors better than others and can't learn some combinations of these cues. Gerbils will work long and hard for sunflower seeds as a reward, and octopuses consistently go for crabs. Early octopus learning researchers found that they couldn't treat a crab as a negative stimulus, when the octopus got shocked for touching it; the octopus just couldn't learn not to try to catch crabs.

Some other constraints on learning that octopuses have might tell us about their anatomy and the way they process information. Wells found that common octopuses can learn by touch and can tell a smooth cylinder from a grooved one or a sphere from a cube. They had much more trouble, though, telling a cube with smoothed-off corners from a sphere, or vertical grooves versus horizontal grooves. They couldn't learn to distinguish a heavy cylinder from a lighter one with the same surface texture. Maybe the common octopus could not use information about the amount of sucker bending to send to the brain and calculate what an object's shape would be, or calculate how much the arm bent to figure out weight. Octopuses have a lot of local control of arm movement: there are chains of ganglia down the arm and even sucker ganglia to control their individual actions. If local information was processed as reflexes in these ganglia, most touch
and position information might not go to the brain and then couldn't be used in associative learning.

While that local control may be true for texture information, the studies on hole drilling for octopuses suggest that they have to use the arm postures and place information for other judgments. We know that octopuses drill into snail and clam shells at particular places, and that these are the right places for the toxin they are going to inject to have the most effect on the prey's muscle. A few years ago, the Seattle Aquarium had an octopus that was so tiny it had never seen a shell it could drill. The first few times, it drilled haphazardly in any area of the shell, quite far from the body or muscle of the clam that it needed to attack. Within a few drills, it changed location to aim over the center of the valve, which was over the heart, and it did this for the rest of its time in captivity. This early learning study needs to be repeated with lots of tiny octopuses, but it's difficult to find them. This learning must be related to position sense, since this penetration of the clam's hard shell is done while the shell is held under the arm web and is out of the octopus's sight.

Detour experiments also have helped us discover whether an octopus can remember where to go. In early studies, octopuses saw a crab through a glass viewing area, then had to move aside, go down a corridor, and turn a corner to actually catch the crab. They could only do this, and with difficulty, if they moved down the corridor keeping in touch with the glass wall and consistently looking toward the crab with one eye. After the more than forty years since the studies, we've learned that common octopuses are usually monocular when looking at possible prey. When an octopus learns about a situation by getting information with one eye, it stores the data in one side of the brain, and by the next day, has transferred the information to both sides of its brain. So when crawling down the corridor, the octopus would only have half of its brain to find its way, following along the wall with one eye and one set of arms.