Literature Review: Punished By Rewards (Part Two)

      Resuming from my previous post on Alfie Kohn's Punished By Rewards, I will now provide examples to show the effectiveness of rewards. In the next post, I will list and articulate the reasons why rewards seem to fail in the long run, and in later posts, provide alternatives to rewards.  

     Imagine a scenario in which a boy began to leave home when his mother calls out him. If you help me clean the kitchen this afternoon, she says, I'll take you to your favorite restaurant tonight. The boy then closes his door and finds a sponge. Then imagine another scenario in which on a teenage girl's list of favorite activities, working on math homework ranks just below having a root canal. Her father announces that if she finished the problem set on page 228 before eight o'clock, he will give her five dollars. The girl promptly pulls out her book.
     What has happened here? Both the boy and the girl complied with someone's else's wishes, engaging in an activity they were otherwise not planning to do (at least not at the moment) in order to obtain something they valued. In each case, one person used a reward to change another's behavior. The plan worked, and that, most of us say, is all we need to know. But let's probe further. Rewards are often successful at increasing the probability that we will do something. At the same time, though, as I will try to show in the rest of this post,  they also change the way we do it. They offer one particular reason for doing it, sometimes displacing other possible motivations. And they change the attitude we take toward the activity. In each case, by any reasonable measure, the change is for worse. The trouble may be not that it doesn't work but that it works only too well, and in this, we pay a substantial price for their success. 

    For whom are rewards effective? Many of the of the early (and highly successful) applications of the principles of behavior modification have involved animals (such as pigeons), children, or institutionalized adults prisoners or mental patients. Individuals in each of these groups are necessarily dependent on powerful others for many of the things they most want and need, and their behavior usually can be shaped with relative ease. Notice that this is not a moral objection; it is a statement of fact about how behavior is easier to control when the organism you are controlling is already dependent on you. In part, this is true because a dependent organism can be kept in a state of need. Laboratory animals are typically underfed to ensure their responsiveness to the food used as a reinforcer. Likewise, in order to make people behave in a particular way, it must be needy enough so that rewards reinforce the desired behavior. People who have some degree of independence will also respond to rewards on occasion, but it is more difficult to make this happen in a predictable, systematic way. 

     For how long are rewards effective? The short answer is that they work best in a short term. For behavior changes to last, it is usually necessary to keep the rewards coming. Assuming your child is reinforced by candy, you can induce him to clean up his room for as long as you keep providing sweets. In practice, however, this raises several problems. What if he becomes satiated with sugar so that the reward eventually stops being rewarding to him? Alternatively, what if his demands to be paid off escalate ( in frequency if not in quantity) beyond your desire or ability to meet them. Most important, do you really want him to help around the house only as long as you have a supply of M&M’s on hand?
     In the real world, even if not in the laboratory, rewards must be judged on whether they lead to lasting change - change that persists when there are no longer any goodies to be gained. We want to know what happens to productivity, or to the desire to read, once the goodies have run out. In theory, it is possible to keep handing out rewards pellets forever. In practice, though, this is usually impractical, if not impossible, to sustain. What's more, most people with an interest in seeing some behavior change would say it is intrinsically better to have that change take root so that rewards are no longer necessary to maintain it.
     If it does make sense measure the effectiveness of rewards on the basis of whether they produce lasting change, the research suggests that they fail miserably. This new should not be shocking; most of us, after reflecting carefully, will concede that our own experience bears this out. However, what is not always recognize is, first, just how utterly unsuccessful rewards really are across various situations, and second, just how devastating an indictment is contained in this fact.

   To start with, let us consider elaborate behavior modification plans such as token economies   (where markets that can be redeemed for privileges or treats are dispensed when people act "appropriately"). Theoretically, these programs should have unusually high prospects for success since they are typically implemented in laboratory-like settings - closed environments with dependent subjects. In the first systematic review of the research done on token economies, conducted in 1972, two avid proponents of the idea stated that:

   "The generalization of treatment effects to stimulus conditions in which token reinforcement is not given might be expected to be the raison d'etre of token economies. An examination of the literature leads to a different conclusion. There are numerous reports of token programs showing behavior change only while contingent token reinforcement is being delivered. Generally. removal of token reinforcements results in decrements in desirable responses and a return to baselines or near-baseline levels of performance".

     Translation: when the goodies stop, people go right back to acting the way they did before the program began. In fact, not only does the behavior fail to generalize to conditions in which reinforcements are not in effect - but reinforcement programs used each morning generally don't even have much effect on patient's behavior during the afternoon.

      One study conducted in a classroom should convey a feel for this kind of research. Over the course of twelve days, fourth and fifth graders were rewarded for playing with certain math-related games and were not rewarded for playing with others. (None of these activities were inherently more interesting than any other). When the rewards started, the kids promptly gravitated to the games that led to a payoff. When the rewards disappeared, their interest dropped significantly, to the point that many were now less interested in them than were children who had never been rewarded in the first place. Researchers concluded that the use of powerful systematic reward procedures to promote increased engagements in target activities may also produce concomitant decreases in task engagement, in situations where neither tangible nor social extrinsic rewards are perceived to be available. 

     In one dieting study, some subjects were promised a twice-a-week reward of five dollars each time the scale showed good news, while others got nothing. Those who were paid did make more progress at the beginning, but then gained back the weight- and then some- over the next five months. By contrast, those who had not been rewarded kept getting slimmer. A similar study published ten years later offered little solace for behaviorists: after a year no difference was found between the payment and nonpayment groups. (Actually, there was one difference: many of those who had been promised money for shedding pounds failed to show up for the final weigh-in). 

     Losing weight and keeping it off are inordinately difficult so it may be unfair to reject pop behaviorism just because it hasn't worked miracles here. The trouble is that it hasn't done much better elsewhere, assuming we are looking for long-term gains. Take smoking cessation. A very large study, published in 1991, recruited subjects for a self-help program designed to help people kick the habit. Some were offered a prize for turning in weekly progress reports; some got feedback to enhance their motivation to quit; everybody else (the control group) got nothing. What happened? Prize recipients were twice as likely as the others to return the first week's report. But three months later, they were lighting up again more often than those who receive the other treatment - and even more than those in the control group! Saliva samples revealed that subjects who had been promised prizes were twice as likely to lie about having quit. In fact, for these who received both treatments, "the financial incentive somehow diminished the positive impact of the personalized feedback." Not only were rewards unhelpful; they actually did harm. 

     At what, exactly, are reward effective?  To ask how long rewards last, and to learn that they rarely produce effects that survive the rewards themselves, is to invite curiosity about just what it is that rewards are doing. Why don't people keep acting the way they were initially reinforced for acting? The answer is that reinforcements do not generally alter the attitudes and emotional commitments that underlie our behaviors. They do not make deep, lasting changes because they are aimed at affected only what we do. What rewards and punishments do is induce compliance, and this they do very well indeed. If your objective is to get people to obey an order, to show up on time and do what they're told, then bribing or threatening them may be sensible strategies. But if your objective is to get long-term quality in the workplace, to help students become careful thinkers and self-directed learners, or to support children in developing values, then rewards, like punishments, are absolutely useless. In fact, as we are beginning to see, they are worse than useless - they are actually counterproductive. 

     In 1961, a graduate student at the university of Kentucky found something she didn't expect. For her dissertation, Louise Brightwell Miller arranged a series of simple drawings of faces so that pairs of nearly identical images would be flashed on a screen. Then she brought 72 nine-year-old boys into her laboratory one at a time and challenged them to tell the two faces apart. Some of the boys were paid when they succeeded; others were simply told each time whether or not they were correct. Miller expected that the boys would do a better job when there was money at stake. Instead, she found that those who were trying to earn the reward made a lot more mistakes than those who weren't. It didn't matter how much they were paid (one cent or fifty cents) or whether they were highly motivated achievers (as measured by a personality test). The discovery left her scratching her head: "The clear inferiority of the reward groups was an unexpected result, unaccountable for by theory or previous empirical evidence" she and her adviser confessed. 

     In a different study, some undergraduate students, instructed to complete a specific construction task, were informed that they could earn anywhere from $5 to $20 if they succeeded; others weren't promised anything. Even though the subjects were older and the assignment quite different, the results echoed Miller's: when the task was more challenging, those who were working for the financial incentive took nearly 50 percent longer to solve the problem. As the 1970's wore on, still more evidence accumulated. Preschoolers who expected an award for drawing with felt-tip pens drew at least as many pictures as those who didn't expect an award, but the quality of their drawings was judged to be appreciably lower. (That rewards can have one effect on quantity and another on quality has been noticed by other researchers too).      Another group of college students took longer to solve a problem requiring creativity when they were rewarded for doing so. And in a particularly intriguing experiment, sixth-grade girls who were promised free movie tickets for successfully teaching younger girls to play a new game wound up doing a lousy job as tutors; they got frustrated more easily, took longer to communicate ideas, and ended up with pupils who didn't understand the game as well as those who learned from tutors who weren't promised anything. 

     By the 1980's, anyone who kept up with this sort of research would have found it impossible to claim that the best way to get people to perform well is to dangle a reward in front of them. As the studies became more sophisticated, the same basic conclusion was repeatedly confirmed. College students exhibited "a lower level of intellectual functioning" when they rewarded for their scores on the more creative portions of an intelligence test.
     A few years later, Teresa Amabile, a leading student on creativity, published two reports that clinched the case against the use of rewards. In the first, young creative writers who merely spent five minutes thinking about the rewards their work would bring (such as money or public recognition) wrote less creative poetry than others who hadn't been reflecting on these reasons for pursuing their craft. The quality of their writing was also lower than the work that they themselves had done a little while earlier. Again, rewards killed creativity, and this was true regardless of the type of task, the type of reward, the timing of the reward, or the age of the people involved. Other investigators, meanwhile, have been looking at people's attitudes toward rewards. Ann Boggiano and Marty Barrett found that children who were extrinsically motivated - that is, concerned about things like rewards and approval they can get as a result of what they do at school - use less sophisticated learning strategies and score lower on standardized achievement tests than children who are interested in learning for its own sake. The reward-driven children do more poorly even when they are compared with children whose scores the previous year were identical to their own.
    

      I have described these studies individually rather than just summarizing the basic finding because without the supporting details of the research the conclusion might be hard to accept - at least until the results appeared so consistently that they had no choice. But before going on the examining the reasons for these results, let us take a moment to sort them out and think about what they imply and why they seem to be so startling. Recall the three questions posed at the beginning of this chapter: For whom are rewards effective, for how long, and at what? We know that some people will do a better job at some things when there's a goody at stake, but few of us have stopped to consider just how limited the circumstances are in which this is true. For whom do rewards work best? For those who are "alienated from their work" according to Deutsch. If what you've been asked to do seems or simple, you might decide to make a real effort only when there is something else, something else outside the task itself, to be gained. For how long do rewards work? Most of the research on this question concerns behavior change, the sort of effect discussed in the preceding post. Virtually all of the studies concerned with performance look at how well people do at a task immediately after getting, or being promised, a reward. In order for rewards to have any hope of boosting performance over a long period of time, we typically have to continue giving them out, or at least holding out the possibility that more will follow.
     

     We come, finally, to the key questions: at what sort of tasks do people do a better job when they are rewarded? And "better in what sense? By now we have already seen enough evidence to guess the answers. Rewards usually improve performance only at extremely simple - indeed, mindless - tasks, and even then they improve only quantitative performance. One of the most influential papers on the topic of rewards (influential, that is, for the very few social psychologists who are specialists in the field) reached the following conclusion based on research conducted up until the mid-1970's: Incentives will have a detrimental effect on performance when two conditions are met: first, when the task is interesting enough for subjects that the offer of incentives is a superfluous source of motivation; second, when the solution the task is open-ended enough that the steps leading to a solution are not immediately obvious. This analysis by Kenneth McGraw provides us with a good point of departure to figure out when rewards are likely to fail. Subsequent investigations, for example, have confirmed that a Skinnerian approach is particularly unlikely to prove useful when it is creativity we are trying to promote. But the research I have described in this section includes enough examples or impaired performance at rather straightforward tasks - or at least a failure to enhance performance at these tasks- that we cannot casually assume it makes sense to reach for reinforcements for everything that doesn't demand creativity.

     "Do this and you'll get that" turns out to be bad news whether our goal is to change behavior or improve performance, whether we are dealing with children or adults, and regardless of whether the reward is a grade, a dollar, a gold star, a candy bar, or any of the other bribes on which we routinely rely. Even assuming we have no ethical reservations about manipulation other people's behaviors to get them to do what we want, the plain truth is that this strategy is likely to backfire. As one psychologist read the available research, people who are offered rewards tend to choose easier tasks, are less efficient in using the information available to solve novel problems, and tend to be answer orientated and more illogical in their problem-solving strategies. They seem to work harder and produce more activity, but the activity is of lower quality, contains more errors, and is more stereotyped and less creative than the work of comparable nonrewarded subjects working on the same problems. In the next post, we will examine why this is all true.  

     If you're interested in the book, you can pick it up from Amazon: Punished by Rewards: The Trouble with Gold Stars, Incentive Plans, A's, Praise, and Other Bribes!