Who Moved My Garden? Spatial Learning in the Octopus

ResearchBlogging.orgSay you're visiting Los Angeles and you have a sudden craving for Chinese food. Since you are only visiting, you might not be aware that nothing is open past, like, 10pm (not even coffee houses), but you get in your rental car and go driving around in search of your Chinese feast anyway. You try hitting up Panda Express, but no such luck. Of course they're closed. You try the neighborhood Chinese restaurant: closed as well. You get back in the car, and think to yourself "maybe the OTHER Panda Express will be open", but alas, it is not. You are ready to return to the hotel and just go to sleep when you think to yourself, "the MARKET is still open." And since you are a culinary genius, you know how to make a homemade version of Panda's orange chicken (I bet you want that recipe, don't you? Don't try looking it up online; you'll find some weak imitations, and I have tried them all). So you head to the market, collect your ingredients, and head back. It's late, but you're motivated enough to cook. Lucky for you, your hotel room has a kitchenette.

Now you're faced with a problem. When you set out on your orange chicken quest, you did not know where you would find those glazed pieces of chickeny goodness. You just drove around. Now you have to find your way back to the hotel, and without relying on GPS or Google Maps, or whatever. Since you're visiting, you don't know the streets at all. You do remember though that your hotel was near a really tall office building. Maybe you could try to find that building.

This is the problem facing many animals, every day. Today, we shall concern ourselves with the octopus.

i-e0ba6afad9720eeaae467d456d02baf2-octopus-thumb-500x388-47437.jpg

Figure 1: The Great Pacific Octopus (Enteroctopus dofleini). PZ said I did not have enough squid. Perhaps this will briefly sate the Pharyngulites?

So one researcher, Jennifer A. Mather, wondered: how do octopuses navigate? Do they rely on chemotactile sensory information, or do they orient towards visual landmarks? Octopuses occupy "homes" for several days or in some instances for several weeks, and when they go out looking for food, they are sometimes gone for several hours at a time. Therefore, they must use some sort of memory to find their way back home.

Many molluscs use trail-following, and follow their own (or their conspecifics) mucus trails. You might expect that octopuses use trail-following as well, since they forage by using chemotactile exploration - at least four different types of receptors on their suckers gather chemical and tactile information as they move along the rocky seafloor.

However, many other species use visual scene recognition to aid in navigation: ants, bees, gerbils, hamsters, pigeons, and even humans, use visual landmarks to navigate around their environments. Since octopuses use visual information to distinguish among different objects, they could use visual landmarks to get home as well.

Field Studies
First, Mather and various Earthwatch volunteers observed the foraging movements of 4 individual octopuses, for a total of 113 complete round-trips. Since only trips which would have required the use of memory were of interest, trips of less than 10 minutes, or less than 2 meters from the starting point were excluded, leaving 60 trips (53%) to be analyzed. The average distance was 9.3 meters, and the average duration was 55 minutes - so memory would surely be required to get home.

i-28aa611cb74d7afdd5d43da8c90b8ab6-octopus trips.jpg

Figure 2: Foraging path of an individual octopus on 7 different trips in July, 1985, near Bermuda.

You will probably notice that, in general, the homebound trips are not retracings of the outbound trips, suggesting that they are not simply following their own trail. Also, the trips were generally in different directions, so they couldn't have simply used chemotactile information from previous trips. Further, it was noted that for longer trips, the octopuses used jet propulsion to move through the water, instead of crawling along with their arms. Since they did not have continuous contact with the ground, it is unlikely that they relied on chemotactile information alone, if at all.

Interestingly, there was a significant correlation between the distance from home at time of capture and the choice of whether or not the captured prey would be eaten out, or brought home for consumption. This suggests that the octopuses may represent their distance from home, or the time elapsed since they left home. (I wrote about this bit of the study last week.)

Then, the researchers and volunteers placed artificial landmarks in front of (but not blocking) the entrance to the octopuses' homes. These landmarks were 6 cm square and 20 cm tall, and painted in alternating patterns of black and white stripes. The landmarks were left untouched for three days, and on the third day, when an octopus was seen leaving home, a volunteer moved the landmark one meter away from its previous location.

Would the movement of this visual landmark affect the ability of the octopuses to find their way home? Not really. This doesn't mean, however, that they don't use visual landmarks - it is indeed possible that they were relying on bigger more obvious natural landmarks (big rocks, cliffs, etc) instead of the smaller foreign landmark. How to address these issues? Controlled laboratory studies.

Laboratory Studies
Four juvenile east pacific red octopuses (octopus rubescens) were housed in one aquarium, but were separated into their own sections.

i-09c8a93b14f56ea217f27cb1e0cf99ae-O-rubescens-thumb-400x394-47444.jpg

Figure 3: East Pacific Red Octopus

Each octopus had, in its section of the tank, a black plexiglass cube with one end open that was used as its "home." For testing, the animal was moved (while inside the "home" cube) into a different tank. The testing tank was circular, and was positioned in the room so that no visual landmarks were available to the octopus from outside the tank. Two pieces of plastic tubing were used as visual landmarks, and were placed directly opposite the opening of the black cube. It took two months for two of the octopuses to become habituated to the testing tank. Two of the octopuses never habituated, and were not tested.

The two pieces of plastic tubing were hollow, 3.3 cm in diameter, and 10 cm tall. The question was whether or not the octopuses could use the tubes as visual landmarks to find food.

First the crabs were simply released into the tank so that the octopuses would become used to hunting and eating the crabs. Once they readily captured the crabs, the crabs were confined to a glass bowl near the plastic tubes. Finally, the crabs would be dropped into the hollow tubes to be retrieved by the octopus, but only after the octopus had grasped the tube - this ensured that the octopus wasn't relying on chemical cues. After this training paradigm which lasted more than two weeks, testing began.

During testing, the two landmarks were systematically moved around the tank with respect to the opening of the home cube. Would the octopuses orient to the visual location of the landmark instead of towards any particular direction, or towards the previous location of the landmark? What about if additional landmarks were added, and the entire visual scene was moved? What if landmarks were switched within the visual scene?

There are lots of conditions here, so I'll just summarize the more interesting ones.

What happens when there is only one visual landmark, and it is moved 90 degrees each day relative to the home cube?

i-85c5fe066439de3a2854c7fec077b4a4-octopus results.jpg

Figure 4: Results for one octopus across 4 days.

The first one was accurate on 6/8 responses, and the errors were 100 degrees, and 180 degrees. The second one was accurate on 7/9 responses; the errors were 100 degrees and 135 degrees. Overall, both animals learned to go to the visual landmark for the food reward, regardless of its location.

What happens when three visual landmarks are available?

octopus results2.jpg

Figure 5: Results for one octopus on 8 different trials. Black square is a black box; small circle is the plastic tube; big circle is a white dish.

This task was only given to one of the two octopuses. The octopus began by orienting towards the larger more obvious black box and then moved from the box to the tube, but by day 3, it had learned to go straight for the plastic tube.

What happens when the three landmarks are moved around with respect to each other?

i-e900dee89d8ec3ec04a5434f38c61b69-octopus results3.jpg

Figure 6: Results for four different trials, on four different days.

The first time the tube was moved, the octopus moved to the box, and then reoriented to the tube. The second time, when the tube was back in the starting location, the octopus initially started moving towards its previous location, but then corrected itself and found the tube. The third day, the octopus explored the box and dish, returned home, and then went directly to the tube. Finally on day four, the first response was to search in the tube's previous location.

Finally, the tube was removed entirely. First octopus went to the box, then the dish, then to the tube's previous location, then back to the box, then it circled the entire tank. On the second day of this condition, the octopus first went to the tube's most recent location, then to the box, back home, and back to the box. The third day was similar.

The apparent strategy the emerges from these conditions seems to be a combination of using the larger stationary items (whose locations are the most predictable, since they didn't move), and and orienting to where the tube had been on the previous day.

What Does It All Mean?
(1) This evidence suggests that octopuses do use visual spatial information. Field studies indicated that the octopuses are not in constant contact with the sea floor, did not retrace their outbound paths to get home, and set out on different paths each day. Given this evidence, it is unlikely that they relied on chemical or tactile information to guide their navigation.

(2) The laboratory tests indicated that they can learn to orient towards visual landmarks, and that they continue to do so even with the landmarks are moved. They could also encode a larger scene consisting of multiple landmarks, and seemed to preferentially orient to larger more obvious objects. The artificial landmark displacement experiment in the field also suggests that octopuses rely on larger, stationary objects (e.g. big rocks) even if smaller objects are more conspicuous.

(3) The systematic errors made by the octopuses in the lab, as well as the distribution of foraging paths in the field, suggest that octopuses maintain working memory for where they have been, and where food has been previously found.

(4) In the field, octopuses' decision to eat their prey immediately or to take it home to eat was partly based on the distance between its current location and home. Importantly, there was no significant contribution of the size of the captured prey to this decision. This suggests that octopuses maintain an internal mental map of their home range (~15 meters diameter), as well as their location within that map relative to home. Other findings, that octopuses may purposefully explore their surroundings, offers supporting evidence for this.

More research is obviously necessary to continue to pick apart these issues, but minimally we have evidence that octopuses can use learn to use visual landmarks to find food or to find home. Do octopuses systematically prefer larger or smaller landmarks? Nearer or farther landmarks? What if visual and chemical or tactile sensors offered conflicting information?

This research is nearly 20 years old, but I chose this paper to write about because it seems to be the first systematic study of spatial learning and navigation in the octopus. Additional research has already begun to address these questions in octopuses and other cephalopods.

Mather, J. (1991). Navigation by spatial memory and use of visual landmarks in octopuses Journal of Comparative Physiology A, 168 (4), 491-497. DOI: 10.1007/BF00199609

More like this

And if you don't blink, you'll notice a fellow blogger's name! ;)

This was my first very small appearance in the scientific literature. I ran all the experiments in the lab. It took quite a while to figure out how to shape the octopus's behaviour in the lab.

"The apparent strategy the emerges from these conditions seems to be a combination of using the larger stationary items (whose locations are the most predictable, since they didn't move), and and orienting to where the tube had been on the previous day."

Even within the two stationary objects, the octopuses usually went for the box first. The box was definitely the easiest to see (black object on white background), and the octopuses seemed to be a bit biased towards it. That makes that the responses to the lower contrast, but rewarded, objects more interesting in some ways.

Great to see this work I was involved in being discussed after... some time. (Oh, who am I kidding? Just another reminder that I am OLD.)

i have a question about studies like this. it boils down to how do you know you don't have a stupid test subject?

for instance, if you take humans and do similar navigation studies, there are people who are bad at navigating. if you happen to get a person who performs poorly, won't it scew the data if you are not aware of it? with a person you can just ask them if they are good at finding their way around. with an octopus you can't. i suppose you could simply not care and use a large sample size, so statistically you would get closer to the behavioral norm. but this study only used 4 octopi. what if one was a particularly poor navigator?

In a sense, we did have animals who were "poor navigators". There were a couple of animals that we were never able to get to learn the task. Never sure why.

Whethere this is a problem depends on which question you're asking.

- What is the species capable of? There, it doesn't matter so much if you have a representative animals that is the "genius" for its species; the question is whether it can be done at all. This is sort of the question researchers will often answer using a small number of highly-trained individuals (e.g, Irene Pepperberg's studies with Alex).

- What is normal for the species? Again, there are a couple of different ways of looking at that question. If by "normal," you want to know a mean and distribution, you need a large sample. If you are looking for whether a trait is generally present at all, the sample size may not be as big an issue.

For instance, if you wanted to know if some humans have hair on their heads at all ever, and you sampled four people at random, it's possible that you might get four bald people and come to an incorrect conclusion. But you're likely to get some idea of something that is common with a small sample, if that sample wasn't biased.

And if you don't blink, you'll notice a fellow blogger's name! ;)

It's true! You're right there in the acknowledgements section. What's the story there? Were you an undergrad RA or something?

Yes, I received a Canadian Natural Sciences and Engineering Research Council (NSERC) summer undergraduate scholarship with Jennifer. I carried on working with her after that. I think I'm in the acknowledgments of a couple of her papers around that time. Some of my own projects didn't get published, but were important in opening doors to a graduate program.

So my scientific career started with a marine animal in the middle of the bald-headed prairie. The absurdities and incongruities have only grown from there.