THE STORY OF learned helplessness began well before that day in 1964 when Steve Maier and I arrived at the University of Pennsylvania. As Steve and I were looking for a project to do together, Dick’s bewildered mention of “helplessness” came back to me, and I talked with Bruce Overmier, Dick’s most senior graduate student, the person who, as best I could ascertain, first discovered it.
For Bruce, two years earlier in 1962, it had been an annoyance to root out, not a phenomenon to investigate. In his first year as a graduate student, he, like many of the brightest kids in the field of learning, wanted to know how the two “basic” learning processes interacted and if they could be unified. This endeavor was a first cousin of Neal Miller’s obsession with finding a single basic process. Bruce’s question was this: If a dog first had Pavlovian fear conditioning in which a tone preceded a shock, then the dog learned to jump a barrier to escape a shock not preceded by a tone, would the tone increase fear and propel the dog to jump even faster?
This was a simple enough question, so Bruce and Russ Leaf, then the senior graduate student, built the equipment and carried out the study. It involved two distinct settings. The first was a comfortable rubber hammock in which the dogs lay, with shock electrodes attached to their back paws. A loud tone sounded, and five seconds later a mild shock—strong enough to make the dog bark but much too weak to do any tissue damage—started, lasting for five seconds. They repeated this sixty-four times over the course of one hour, ample Pavlovian conditioning trials to make the dog fear the tone. This first part went according to expectations.
The second part introduced the big, unexpected problem. Russ and Bruce built a shuttle box, a six-foot-long, two-chambered room with steel grids on the floor and a low barrier between the two sides. In the shuttle box, dogs learned easily to avoid shock: the lights dimmed and five seconds later the grid floor was electrified—again with a mild-moderate shock, just strong enough to provoke a jump over the barrier to escape. The idea was to turn on the tone from the earlier session in the hammock and see if it made the dogs jump faster.
But the experiment fizzled. The dogs conditioned in the hammock did not jump the barrier. They just lay down and did not try to escape the shock. This was unprecedented and enormously frustrating for the investigators.
This I had to see. So Bruce took Steve and me up to the lab on the third floor of the decrepit Hare building, just around the corner from College Hall, to show me the dogs. Indeed the dogs looked helpless—whatever that might mean in scientific terms. Bruce also showed me the two-inch-deep gouge in the sickly green wall where a furious Russ Leaf threw his clipboard in frustration.
What Bruce and Russ saw as an annoyance,1 Steve and I saw as the phenomenon, one of much farther-reaching importance than unifying Pavlovian conditioning and instrumental learning. Helplessness in humans is ubiquitous and the source of endless suffering. Could we capture it in the laboratory? Could we discover what causes it? Could we discover how to reverse it? Could we discover its brain physiology? Could we discover how to prevent it?
This was why I became a psychologist.
FROM THE BEGINNING I thought the phenomenon looked like helplessness. But what, we puzzled, were the elements of helplessness? How did it come about? How could we test for it? What were its consequences? How did it relate to mental illness? How could we begin to do serious science on the mere intuition that the dogs had become helpless?
I met Steve Maier on my first day in graduate school at Penn, and we agreed to collaborate on the spot. He became a serious neuroscientist and upended our theory of learned helplessness. Steve is pictured here at the time he made his major discoveries that overturned our original theory. Photo courtesy of Steve Maier.
We began by defining it. Central to our intuitive notion of helplessness was the belief that nothing one does matters. This was subjective helplessness, but what, first, were the conditions of objective helplessness that might lead to such a belief? This was a thorny question, and its solution ran afoul of the most sacred premises of the learning theory that ruled academia. Formally, an animal or a person is objectively helpless about an important outcome, such as shock offset, if the probability of that outcome, given a response, does not differ from the probability of that outcome in the absence of that response. When this is true of all responses, objective helplessness exists.
Notice—as every learning theorist had previously failed to—that all Pavlovian conditioning is helplessness: no response of the animal made any difference to the outcome. The food came whether or not the dog salivated. The shock followed the tone whether or not the rat’s heart rate rose. Pavlovian conditioning entails objective helplessness by definition, and it seemed never to have dawned on anyone that animals might learn this and that they might find this more urgently important than the Pavlovian part of the experiment. A strong theory, such as Pavlov’s, tells scientists what to look at and what to ignore, and around Pavlovian conditioning there descended a hegemony of silence about what to ignore.
Once you digest this unpacking of “nothing one does matters,” helplessness seems straightforward—there is no relation between what happens to the dog and what the dog does. So we proposed that animals (and people) could learn that they were helpless, and for that to happen, the animal had to “detect” this lack of relation and so must be able to learn that outcomes like shock offset were independent of its responses. This was a radical suggestion for the 1960s. The learning theorists of that era, the grandchildren of early behaviorism, were stubbornly “associationistic” and militantly anti-mentalistic. Cognition was to be shunned, and the underlying program was to show that the “illusion” of cognition entailed nothing more than simple stimulus-response associations. Animals can learn only pairings—for example, a response paired with shock, strengthening the association (acquisition), or a response paired with no shock, weakening the association (extinction). These pairings are the only two kinds of “magic moments” that can produce learning. It seemed to us that the integration of two conditional probabilities—the probability of shock given a response and the probability of shock in the absence of that response—was an act that required brain power rather than mere behavioral conditioning. For this reason we called our theory “learned” helplessness rather than “conditioned” helplessness, in the process rejecting the simple premises that both Pavlovian conditioning and instrumental learning held so dear.
Emblematic of the tension between learning theory and cognitive theory was a standoff between Dave Williams, a young radical behaviorist professor, and me in the first-year learning proseminar.
“So you think, Marty, that animals have a rich mental life?” Dave asked, as if mockery should end the argument. I did indeed think that higher mammals were richly cognitive and that the ability to form expectations about the future had given us an edge in the never-ending competition to survive. But these beliefs were so out of keeping with behaviorism that one either shut up or one endured mockery. I didn’t shut up, and neither did Steve.
All of this made us outcasts on our home turf by putting us squarely in the camp of what would soon be called “cognitive” psychology. Two winds were already blowing in the gathering storm of the cognitive revolution. The first was Jean Piaget’s sound scientific observations on the mental processes of children. The second was Noam Chomsky’s devastating review (1959)2 of B. F. Skinner’s long-awaited volume (1957)3 allegedly demonstrating that language could be understood in strictly behaviorist terms. Chomsky, a rebellious linguist and talented polemicist, argued that the essence of language was not reinforcement—the repetition of old verbal habits—but generativity: speakers could produce and understand entirely novel utterances, such as “There is a purple Gila monster flying through the kitchen,” never before encountered by anyone. Learned helplessness, arguing that animals were cognitive, was the third wind, and the fourth wind, the one that transformed this into a hurricane, got started in 1965 at Penn.
Ulric Neisser, on sabbatical from the Brandeis University psychology department, was spending 1965–1966 at Penn writing his great book Cognitive Psychology.4 He offered us a graduate course, which consisted of discussing his working manuscript. Dick, as we all called him, argued that behaviorism got its sea legs by somehow convincing psychologists that replicable, objective science on mental life was impossible. New experiments showed that they were wrong, and this kind of work justified a new science of the mind. Just one example suffices.
Briefly flash up on a screen a brightly lit ten-by-ten matrix of random letters. Turn it off two seconds later. It disappears from the screen, but the subject still has an “afterimage” of the matrix before his eyes. Now flash an arrow where the third row of the matrix used to be, and ask the subject to read off the letters still in his afterimage. The subject gets seven of the ten right, and this quantifies how much information is still there in the mind’s eye two seconds after the objective stimulus is gone. How long does the afterimage last? Instead of waiting two seconds, vary the interval, and discover the rate at which the afterimage melts to nothing. This is a rigorous measure of the rate of decay of a mental image.
The scientist can even tell precisely how bright the afterimage in the mind’s eye is. During the interval, flash up masking lights of differing intensities and discover the brightness at which the afterimage is exactly masked. The duration, the amount of information, and the brightness of a mental image can be exactly measured. So very fine, quantitative science can be done on mental life.
Dick published his blockbuster book the following year, and the cognitive revolution—which is still with us in 2018—began to sweep behaviorism away, just as behaviorism had swept mental life away at the end of William James’s life.
Steve and I, deeply impressed, were also in the midst of testing whether animals were cognitive. How could we find out if it was the lack of contingency between response and shock and not the shock itself that produced the profound failure to escape in the shuttle box? How could we find out if animals actually learned that nothing they did mattered?
To assess this question, we had to isolate the noncontingency from the shock. So three groups were needed. One group got escapable shock (ESC), where shock offset was contingent on the animal’s response. So in our first helplessness experiment,5 these dogs learned in the hammock to press a panel with their noses to turn off the shock. The second group was yoked to the ESC group, getting exactly the same duration, intensity, and pattern of shocks, but their responses had no effect on the shock. In this inescapable shock group (INESC), shock offset and all the animal’s responses were noncontingently related. A third group (Zero) got no shock.
The next day the dogs were tested in shuttle box escape. We ran twenty-four dogs through our new “triadic” design and found that all the dogs in the ESC group and all the dogs in the Zero group readily learned to escape in the shuttle box, but five of the eight dogs in the INESC failed altogether to learn to escape. Importantly the ESC and the Zero group learned to escape equally well. We concluded that the animals in the INESC group learned in the hammock that shock offset was independent of their responses, and when shock occurred the next day in the shuttle box, they expected that its offset would once again be independent of their response. This expectation undermined their trying (“response initiation”) to escape. They had learned helplessness.
Our first publication on helplessness, outlining our cognitive theory, was, to our surprise, published as the lead article in the prestigious but archconservative Journal of Experimental Psychology.6 The reviews contained only one criticism: “paralyse is usually spelled with a ‘z.’” We briefly harbored the illusion that learned helplessness would be well accepted, and it would be smooth sailing from there, but we were wrong. The learning theorists, sensing that the rising storm would soon pose a mortal threat, closed ranks against cognitive theorizing in all forms, particularly in animals. There was a hostile encounter at the Princeton University conference where Steve and I first unveiled the theory to the major learning theorists:7 Dick Herrnstein, B. F. Skinner’s right-hand man and articulate gunslinger, retorted, “You are proposing that animals learn that responding is ineffective. Animals learn only responses—they don’t learn that anything.” Animals for the behaviorist are stimulus-response machines, not creatures that can ever learn propositions.
We defined a dimension that we called “control over outcomes,” with control being present whenever the probability of an outcome, given a response, differs from the probability of that outcome in the absence of that response. Clearly, the escapable subjects controlled an aspect of the aversive event (when the shock went off), and the inescapable subjects did not. This was exactly why we developed the triadic design and included the escapable subjects as a control group as well as the zero group, because it isolated the element of uncontrollability: if failure to escape in the shuttle box was caused by learning uncontrollability, then this failure should not occur if uncontrollability was removed but the shocks stayed identical. That is, we assumed that uncontrollability was the active ingredient in producing passivity and that escapable subjects were later escaping normally in the shuttlebox because they lacked this critical learning ingredient.
We found out, almost fifty years later, when Steve finally unraveled the brain processes, that our theory was 180 degrees off. Steve discovered that helplessness is not learned; rather it is mammals’ default response to prolonged bad events. What is vouchsafed to mammals is learning that bad events can be controlled. One of the privileges of doing science for fifty years is that it is long enough to find out that you were wrong, and I will tell you the story of Steve’s discovery of the Hope Circuit in Chapter 28.
Much of the work Steve and I did in graduate school centered on testing the theory of learned helplessness. But I became a psychologist not to test theories but to relieve suffering, and animal experiments really only mattered to me if they could lessen the amount of misery on the planet. I was sure that what we had seen in our dogs was about something larger. So I began to study abnormal psychology under Jim Geer, a behavior therapist newly arrived at Penn as an assistant professor. I told Jim about our dogs, and he came up to the lab to see them.
“Those dogs are depressed,” Jim told me.
“What is depression?” I asked Jim. I spent much of the next ten years trying to answer my own question.
So in addition to theory testing we did two kinds of experiments anticipating the clinical applications of learned helplessness. First, we asked about prevention—we called it “immunization,” influenced by the Salk and Sabin vaccines that had recently put an end to polio. If our dogs first learned that they could control shock and then experienced inescapable shock, would they not become helpless? Immunization worked: dogs that first learned to turn off shock in the hammock by pushing a panel with their noses, then had enough inescapable shock to make naive dogs helpless, later easily learned to escape in the shuttle box by jumping the barrier. These dogs, we believed, expected that they could control shock and so did not give up. They were immunized against helplessness.
We also worked on “therapy.” Once a dog was helpless in the shuttle box, what could we do to cure him? We took inspiration from “exposure” work in behavior therapy. Spider-phobic people, for example, who are forced to endure being in the presence of spiders until they discover that nothing bad happens, are cured. So we decided to show the helpless dogs that they had control over shock. We dragged them across the shuttle box and out of the shock that they were passively enduring. We showed them that getting to the other side worked. After a few draggings, the dogs perked up and began to jump on their own. This was true of every dog, a 100 percent cure.8
Ethics
There was a snowy-white, long-haired dog in the laboratory. I spent a lot of time with him. I loved him, and he loved me. I wanted to adopt him, but Steve and I weren’t allowed pets in our apartments. So after the experiment we took him in Steve’s aging Plymouth Valiant to Fairmont Park, the Manhattan-sized park in the middle of Philadelphia, and we set him free. Two weeks later he reappeared in our laboratory, captured once again by the people who rounded up stray dogs and sold them to laboratories. I do not know his fate, but I have dreamed about him at least once a month for the rest of my life.
Running dog experiments was a harrowing experience for us. We were both dog lovers, and had we not found ourselves in a laboratory in which dog experiments were already underway and unquestioned, we would not have chosen this path. As soon as I could I stopped experimenting with dogs; I ran rats and then people in helplessness experiments, with exactly the same pattern of results as with dogs. I have not run an animal experiment in more than forty years, and I have thought long and hard about the practice.
There are three kinds of issues that bear on the “ethics” of animal experiments such as ours. The first is “sentimental.” It feels bad to do these experiments. Unlike some scientists, I never habituated to feeling bad. I cringed then and I still cringe remembering them. I still dream of the white dog, with love and with guilt.
The second issue is “moral.” Peter Singer, a critic of learned-helplessness experiments, has put it well.9 To wit: all life is sacred, and a civilized morality extends the human moral circle to animals. Animals are equally entitled to our sympathy and protection as other human beings are. Singer considers it “speciesism” to believe that humans deserve special rights. In this view, even if the shock was mild to moderate in our experiments (and I tested it on myself, producing a sudden startle but no pain), it is always and absolutely wrong to shock animals.
I strongly disagree with this kind of moral absolutism. Inflicting pain on animals (or people) is wrong, but not absolutely wrong. This is but one value in a complex world of clashing values. Curbing the epidemic of depression, suicide, and posttraumatic stress is another value. Markedly decreasing the tonnage of human suffering in the world matters more to me and to most other people as well. These are difficult choices, particularly when the cringe factor is strong and one cannot see clearly into the future. The progress that was made over the next decades in curing and preventing mental illness as a result of our studies—which I discuss in later chapters—tells me that we saw into the future with considerable accuracy.
The final ethical issue is “scientific.” All animal experiments that try to make inferences to humans must tangle with the problem of external validity. This is a crucial, neglected, and really hairy issue. Experimental psychology’s rigor, its internal validity, had attracted me to the field. A controlled experiment is the gold standard of internal validity because it discovers what causes what. Does the fire cause the water to boil? Turn on the fire, and the water boils. Without the fire (the control), the water does not boil. Do uncontrollable bad events spur tumor growth? Implant tumors in rats and then give one group of rats inescapable shock, give another group identical escapable shock, and compare them to a group that receives no shock. Rats that get inescapable shock grow a cancerous tumor at a greater rate; hence, inescapable shock causes tumor growth in rats.10
But what does this tell us about the causes of cancer in human beings, and what does that tell us about how helplessness influences cancer in human beings? This is the problem of external validity. When laypeople complain about psychology experiments with the gibe “white rats and college sophomores,” it is external validity that is in dispute. It may seem simplistic, but it is a profound complaint. Human beings differ in many ways from the white laboratory version of Rattus norvegicus. Inescapable shock differs in many ways from finding out that your child drowned in a boating accident. The tumor implanted in Rattus norvegicus differs in many ways from the naturally occurring tumors that afflict human beings. So even if the internal validity is perfect—rigorous experimental design, the exactly right control group, large enough numbers to insure randomization, and impeccable statistics—we still cannot infer with confidence that this illuminates the effect that uncontrollable bad events will exert on cancer in people.
I have come to think that establishing external validity is an even more important, but much more nettlesome, scientific task than establishing internal validity. Academic psychology requires all serious psychology students to take entire courses on internal validity. These “methodology” courses are entirely about internal validity and almost never touch on external validity, which is often passed off as mere philistinism about science on the part of ignorant laymen.
Hundreds of professors of psychology make their livings teaching “scientific” method and statistics, the trappings of internal validity, but no one makes a living teaching about external validity. Unfortunately, public doubts about the applicability of basic, rigorous science are often warranted, and this is because the rules of external validity are not clear.
As you will see in later chapters, I went on to work for years asking if the helplessness experiments in dogs and then rats and then people had external validity and if they illuminated real helplessness in real human beings.
They did.
So learned helplessness in dogs was the choice I made, and if it were offered to me again, I would steel myself and make the same choice.