In the wild, ravens’ food is never found dangling on anything that resembles a string. String-pulling could therefore not be hard-wired into the innate behavioral repertoire by natural selection over millions of years of foraging experience. The behavior might, of course, be learned in incremented steps by tedious repetition where you reward the first step that contributes to the correct solution, then you reward the second step, and so on until the bird ultimately strings together a dozen or more steps. My captive ravens had never before even seen string. I had a unique opportunity to do a useful experiment. I doubted they could perform the task, but if they did, then innate programming or learning could both largely be excluded, leaving either insight or random chance as the prime contender for an explanation.
As I contemplated the problem of how a bird would get food dangling from a string, I saw a unique opportunity for testing insight, a mental capacity that defines at least one of several proposed kinds of intelligence (Gardner, 1998). First, food should give a strong emotional motivation to induce a bird to try to get the food the easiest, quickest way. The bird might try to sever the string, fly at the food directly, or attempt to knock the food off. Yet, given a secure string, there was only one sure and easy way to get the food—pull it up in successive steps, hanging on to each pulled-up loop of string.
Pulling up food dangled on a string involves many steps that must be executed in a precise and nonarbitrary sequence. The bird must 1) perch above the dangled string that is attached to the food, 2) reach down below the perch, 3) grasp the string with its bill, 4) pull the string up and over the perch, 5) place the pulled-up loop on the perch, 6) lift one foot, 7) step on the loop of pulled-up string with the lifted foot, 8) press down with the foot hard enough to prevent string slippage, 9) release the bill’s hold on the string, but only after the foot is already firmly pressing the string down on the perch, and 10) repeat the whole sequence a variable number of times depending on the length of the string and the amount of slippage of the string.
I arbitrarily chose a string length of two and a half feet because I felt it would present the birds with a sufficiently challenging problem. The process would have to be precisely repeated at least five to six times, or perhaps many more times if the string were to slip. In all, dozens of discrete steps had to be assembled into one precise sequence needing constant updating as it progressed. Any one step by itself could of course be learned and/or innate, but the critical behavior was not any one individual step or steps. Instead, it was the placing of these steps into a unique sequence that would solve a unique problem. It was practically implausible that the correct sequence of dozens of discrete steps would emerge by random chance.
The beauty of the string-pulling test was that 1) it greatly reduced the possibility that the solution could be arrived at by random chance, 2) there was no plausible genetic programming that could have coded this very specific unnatural behavior because there is no conceivable reason for it to have evolved in the wild, and 3) my ravens had been hand-reared in captivity in an aviary and had no experience in string-pulling. They had not even seen string, hence, I knew they could not have had prior opportunity to learn the behavior.
My ravens, I presumed, would not be able to get meat by pulling it up on the string. If they did not, so what? Nothing would have been lost. On the other hand, if they did it from the start, a lot would be won. Invoking Occam’s razor, the rule in science that favors the simpler theory over the more complex one, a successful pull-up would suggest that the birds could solve a problem. That’s proof of insight and if one grants it is a difficult problem, then its solution is also a measure of intelligence as regards that problem. Furthermore, solving the problem by insight presupposes consciousness. And all I needed for this was a piece of string and a slice of meat—not exactly an impressive investment. It was worth the try.
When I first thought about doing the experiment, I recognized two complications. First, ravens are highly temperamental animals. They would be wary of the string. In one test out in the field, I hung a piece of meat on a white piece of twine from a branch next to a rock-solid frozen cow carcass where more than fifty wild ravens were feeding at temperatures near minus 25 degrees Fahrenheit. They had to work long and hard to chip off tiny bits of meat from the carcass, and every dawn the crowd came and all chipped for hours. Any loose piece of meat was a valuable prize, to be taken instantly and much fought over. So here was suddenly a morsel of choice food dangled next to them. What happened?
I was as usual well hidden in my blind of spruce and fir. I had put up the meat while it was still dark. At dawn, the birds came and instead of quickly descending to feed as they usually did, they stayed up in the trees, making angry, rasping alarm calls such as they make to strange, frightening phenomena. Only about an hour later did one of them descend to the cow, leading the others in. They looked at the meat on the string as if it were an apparition. Not one went near it. I left it there. Two days later, it was still there.
I hoped that my aviary birds would be less spooked by such a strange thing, but I expected that they would be shy, since they had never seen or experienced food on string. If they did figure out a solution to reach the meat, they still might not demonstrate their insight because of fear of approaching the string. Ideally, I would test them one at a time, but that meant catching them and building another aviary to isolate them, which would cause much disruption, money, and time. Isolated aviary birds would likely stay away for days from something as unusual as food dangling on a string. The ravens would be much calmer if they remained together as a group. There was a negative aspect to working with a group of birds, because if one pulled up food, the others might simply copy the first, and/or chase away all others. Nevertheless, the first bird I tested had none other to copy, and I was interested in the phenomenon as such rather than in how many individuals could or could not do it. It made no difference to me whether ninety-nine percent or one percent of the population of ravens could solve the problem. If problem-solving—whether you choose to call it genius (in ravens), insight, or intelligence—proved to be present even in one bird, then it exists. Even if rare, it is no less interesting a phenomenon.
Hard salami was important for my test. I first offered pieces out of my hand to make sure they would want it; they ate it eagerly. I knew if I had used soft meat and a bird had flown at the suspended prize, grabbing it with its bill, it might have torn off a small piece. Being rewarded, the bird would be predisposed to try the same maneuver again and again; it would not find it necessary to try anything different. Hence my test would fail.
The salami was three months old, having dried in the refrigerator to a leather-like consistency. The ravens would not be able to tear it off by flying at it, grabbing it with their bills, and then dangling from it by holding on to it with their bill, behaviors I was sure they would try if they tried anything.
I notched a slice of salami and tied it onto the end of tough woven string, then tied that onto one of the horizontal perches in the aviary so that the prized food item was suspended about six feet off the ground and two and a half feet down from the perch. I rushed back into the house to my desk by the window to watch.
All of the birds looked at this new thing. The dominant pair came closer and tilted their heads and stared at the salami. They looked at the string wound round the perch, then hopped up and down on the perch doing nervous jumping jacks, as they would do in front of a feared carcass. The string and the salami jiggled. After a while, and more looking, they cautiously approached. One of them pecked at the loop of string tied around the perch and quickly jumped back. They kept craning their necks to look down. One gave the string a few tugs as if trying to rip it off the branch. The meat jiggled some more, but the string was tough and did not break. The birds then seemed to lose interest. I went back out to remove the string with the salami. I would have them try it again some other day. “Just what I expected,” I thought. “No way will these birds get that meat.”
When I put the meat out the second time the two birds again eyed it warily, but looked less nervous than the first time. Abruptly, one of the birds, Matt, flew up to it, and to my utter amazement performed the entire sequence with only minor fumbling. I shouted for joy and pounded on the window to startle him so he would drop the food. I wanted to make sure his reward was only mental. Not having eaten, would he go through the entire sequence again right off?
This bird knew very well what he had done. After I shooed him off, he returned within several seconds, pulling the meat up in great haste. Again and again (six times) I chased him off before he had a chance to eat. Interestingly, he never tried to fly off with the salami after pulling it up. Furthermore, he always dropped it even when I rudely startled him and later when I literally had to push him off his perch after a meat pull-up when he held the meat in his bill. In contrast, when the birds get a piece of loose meat they always fly off with it.
After the sixth repeat of the pull-up behavior, I knew it had not been a fluke. Matt really did know how to pull up meat. As a reward after all those “empty” repeats, I finally allowed him to eat salami. My reward was an adrenaline rush, and the chance to do a series of experiments that built on these original observations. I was hooked. I was convinced that I had blundered upon an insight for testing insight in ravens.
Eventually, I saw other naive ravens do the pull-up of the same length in as little as six minutes from the time of first presentation of the meat on string into the aviary, and after as little as thirty seconds after they first contacted the string. Given that many of the birds were shy of strings, the time required to pull up the meat successfully probably greatly underestimates the amount of time it takes them to do the task, after they see meat on string. The presence of the other birds turned out to be a big problem (which I solved in another set of observations with another group of birds) because whenever a dominant bird who knew how to pull up meat saw a subordinate near meat on string, it tried to chase the bird off.
I eventually tested five different groups of ravens and two crows. In no case did any naive birds ever show any interest in pulling up, or even approaching a string without food attached to it. In all groups of ravens, but not the crows, several individuals performed the whole sequence with little or no fumbling right from the start. Two patterns were seen. In one pattern, the “direct pull-up,” the bird stayed in place to pull up successive loops carefully, stepping as the loops pulled up. In the other, the “side-step,” the bird pulled string laterally onto the perch before stepping on it, then pulling up again, and so on. On the other hand, a group of four three-month-old ravens, like the two crows, were incapable of reaching the meat, although they showed no fear.
To me, the most convincing point of the experiments that showed insight (because it excluded both observational learning and trial-and-error learning) was that those birds who didn’t pull up meat, but who seized meat on a string that I or another bird had pulled up, tried to fly off again and again with that meat. Of course, since it was still attached to the string, it was rudely yanked out of their bills. They didn’t seem to catch on until at least six trials. Those birds who did pull the meat up, on the other hand, never once in thousands of trials flew off with it so long as they were left on their own. I had difficulty shooing them off the perch after they had pulled up the meat, and when I did succeed in getting them to leave, they almost always dropped their piece of meat before flying off. The ravens that had pulled up the meat acted as if they knew, from the very first time they pulled it up, that to try to fly off with the meat would involve having it ripped out of their bill in flight. The significance of the remarkable behavior of not flying off was that it was a new behavior that was acquired without any learning trials. They acted as though they had already done the trials. The simplest hypothesis is that they had—in their heads.
I next demonstrated experimentally that one behavior—choosing the correct string when two were side by side—could be the result of two different mental concepts. In the setup for this experiment, the birds that were experienced in pulling up meat were first exposed only to one string with meat. In the competition among each other they all rushed to be first at that string, and they pulled quickly. That is, I trained them to expect meat every time they pulled up string, hence looking at the string closely before pulling it up became unnecessary if not counter-productive. Then I provided two side-by-side strings, one holding a rock, the other, meat. There was again competition among the five birds to be first to pull up the meat. In this situation, there was as before much haste and some birds initially made mistakes; they rushed to the strings and yanked on the first one they came to. However, if that was the wrong one, they quickly realized their mistake. They dropped that string without pulling it up, looked again, and pulled up the one with meat. That is, I had now retrained them to look. As one might expect, after a few trials they learned to look before they gave their first yank. A choice was now necessary, and they learned to contact only the correct string. Then in the test, I crossed the two strings so that a raven perched directly above the meat and pulling on the string below its feet would now (in contrast to all its previous trials) end up pulling on the string with rock, not food. Conversely, to get meat, the raven now had to perch above the rock and pull the string below its feet, a novel setup contrary to its previous string-choice training experience.
In the test, three out of four proficient string-pulling ravens first contacted the wrong string—the string attached to the pole directly above the meat. That by itself was neither surprising nor interesting, because that choice previously had always been the correct one; but what was interesting and amazing was that now they showed no evidence of learning to correct their mistakes. In dozens of trials, they continued to yank first on the wrong string; i.e., the one over the meat, switching only after they had seen their mistake; i.e., the rock jiggled. That is, consciousness of what they thought they knew took precedence over trial-and-error learning, which was glacially slow even for this one extremely simple task.
Proof of conscious involvement was shown by one bird who was correct from the beginning. Before the test, that one bird had shown identical behavior to the other three. That is, it pulled, as trained, only on the string with meat, the one directly above the meat. In the test, this bird immediately and consistently did a novel thing. It pulled the string over the rock, the string to which the meat was attached. Rather than first contacting the “string-above-meat” as the other three had done and continued to do without correcting themselves, this bird had pulled on “string to which meat is attached” on the very first and all subsequent trials. Of course, the birds could not use words, but words are unnecessary for thinking as such. (If the ravens had evolved to communicate to their followers how to pull up string or some other useful tasks, then they would of course have had to evolve the capacity to use words, and to communicate using them they would then need the ability to think with them as well.)
In subsequent tests, birds that were proficient in string-pulling were given food on a new string of a different color, texture, and thickness, with which they had never been rewarded, versus a rock on the same string on which they had always been rewarded. If they simply had been conditioned to pull on brown twine, for example, then they would choose it above previously unfamiliar green shoestring, even when the food was attached to the shoestring, which they had never pulled up before. What did they do? They all chose the new string that they had never seen before, much less been rewarded from, right on their first trial. They ignored the familiar twine that had always been associated with food. In summary, they knew the solutions to several new tasks without any overt trials. Given a choice, they attended to what is relevant in preference to what they had been trained, and what they had in mind could take priority over what they experienced. In conjunction with all the other prior observations that they can keep track in their mind of what they no longer can see, I conclude that they experience some level of cons
ciousness, and use it for insight to make decisions. Whether that is “intelligence” is subjective; but according to most people it is.
I wrote up the data, providing thoughts on what it might mean, and submitted the manuscript to a journal that is reviewed by other scientists whose names remain secret to all but the journal editor. The editor then considers the comments of the reviewers and accepts or rejects the paper, generally asking the researcher to consider the reviewers’ comments in the revision if it is to be published. Constructive comments on one’s work are precious, because they often catch one’s errors and oversights. I anticipated few problems. Since I described something novel, I felt sure that my paper would be received eagerly and rushed into press. However, this manuscript took a different turn.
One of the reviewers felt that I needed to examine the evolutionary precursors of the behavior, and also pointed out that “Freud had shown much mental work to be unconscious and that much of human insight is unconscious.” Sure. Okay. Does that then preclude further research? What is mental work anyway? What evolutionary precursors of string-pulling? The editor rejected the paper, but offered to consider it again after a new submission, subject to another round of reviews. I rewrote and resubmitted.
Mind of the Raven: Investigations and Adventures With Wolf-Birds Page 36