Drawing for rewards
Extrinsic rewards can "crowd out" intrinsic motivation
Imagine you’re a kid, and you just ate dinner with your family. You’re in a good mood, so you decide to voluntarily wash the dishes. As you stand up, your mother says, “If you do the dishes, you can watch a movie before bed!”
What a great deal — you were about to do the dishes anyway! But what does your mother think? You could explain your original intent — “Just so you know, I was about to do the dishes anyway” — but she might not believe you (or worse, rescind the offer), so you don’t say anything. As a result, she probably thinks that you’re only in it for the movie. And interestingly, as you wash the dishes, you notice that you’re feeling the same way. Tomorrow night, as you finish dinner, you’re not so eager to voluntarily wash the dishes.
The overjustification hypothesis
This is an example of the overjustification hypothesis. Lepper, Greene, and Nisbett defined it in a 1973 paper as “the proposition that a person’s intrinsic interest in an activity may be undermined by inducing him to engage in that activity as an explicit means to some extrinsic goal.” You initially wanted to do the dishes, but your mother offered a reward, so in the future, you might not want to do the dishes as much.
The first paper to demonstrate the overjustification hypothesis was Edward Deci’s 1971 paper. In the first round of his experiment, the researchers asked college students to do some puzzles. In the second round, they paid half of the students based on their performance on some more puzzles. In the third and final round, everyone did puzzles but nobody got paid.
To measure the students’ intrinsic motivation, the experimenters did something sneaky: during each round, they left the room and told the students they could do whatever they wanted, and they secretly noted the students’ behavior. The results were consistent with the overjustification hypothesis: the paid students spent less time on the puzzles in the final round than they did in the first round, while the non-paid students spent more time in the final round than they did in the first round. Although paying the students “worked” in the short term, it backfired in the long term.
Deci only had 24 students in his experiment, so Lepper, Greene, and Nisbett sought stronger evidence of the overjustification hypothesis. They also wondered what would happen if some subjects received a reward without expecting one. The hypothesis was that this wouldn’t decrease intrinsic motivation, because if the suspects don’t know about the reward, it couldn’t overshadow their intrinsic motivation to do the activity.
Lepper, Greene, Nisbett (1973)
In their experiment, Lepper, Green, and Nisbett recruited preschoolers who seemed interested in drawing and split them into three groups:
Expected-award: These subjects agreed to do some drawing activity in order to obtain an award (a certificate with a gold seal and ribbon), which they received after doing the activity.
Unexpected-award: These subjects did the same activity and received the same award, but didn’t know about the award until after they had finished the activity.
No-award: These subjects did the same activity, but they neither expected nor received an award.
One to two weeks later, the experimenters secretly observed how much time the subjects spent drawing in their natural classroom settings. Again, the results confirmed the overjustification hypothesis: the subjects in the expected-award group spent less time drawing than those in the other groups, and this time, the results were statistically significant.
Moreover, compared to before the experiment, the expected-award group was significantly less interested in drawing, while the other groups were slightly more interested. Finally, the experimenters asked three judges (blind to the experimental conditions) to rate the drawings from the experiment on a scale from 1 to 5. Unsurprisingly, the expected-award group had significantly worse ratings: 2.18 against 2.85 (unexpected-award) and 2.69 (no-award). In summary, students who expected and received an award drew worse, and later, drew less.
The authors point out that these results prompt us to reconsider “traditional classrooms where systems of extrinsic rewards — whether grades, gold stars, or the awarding of special privileges — are applied as a matter of course to an entire class of children.” They refer to prior work that collectively suggests “the schooling process seems almost to undermine children's spontaneous interest in the process of learning itself.”
Lepper and Greene (1975)
Given the potential implications of the overjustification hypothesis (like abolishing grades), Lepper and Greene sought to replicate their findings in another paper. This one examines the impact of rewards as well as surveillance on children’s intrinsic motivation. The activity was solving puzzles instead of drawing, and the reward was the opportunity to play with some attractive toys.
The experimenters split the children into four groups: yes/no surveillance, and expected/unexpected reward (ER/UR).1 The ER subjects were told that they'd get to play with the toys if they did "a good job," while the UR subjects were not told about a reward. After doing the puzzles, all subjects were told they did a good job. To the ER subjects, the experimenter added that they had earned the chance to play with the toys. In contrast, to the UR subjects, the experimenter added “simply and in a noncontingent manner” that they could play with the toys.
The surveillance part of the experiment involved sitting some subjects in front of a camera and telling them that the experimenter would watch them do the puzzles through the camera.
One to three weeks later, two observers (blind to the subjects’ experimental conditions) secretly measured how much time the children spent voluntarily doing the experimental activity. Again, despite having a different activity and reward, this study found evidence of the overjustification hypothesis: expecting a reward and knowingly being surveilled both correlated with decreases in subsequent interest.
Lepper and Greene end with a warning for teachers: although extrinsic pressure is sometimes effective, it can create a negative, self-fulfilling cycle. If the student gets rewarded for doing an activity, then naturally, the teacher might attribute the student’s behavior to the reward. So the teacher “maintains that external pressure,” and this maintenance “may ultimately convince the child or worker himself that he is motivated by the external pressure, making him less likely to engage in the activity in the later absence of that pressure.” In other words, why wash the dishes if there’s no movie?
It’s hard to break out of this cycle, because “the mere existence of such controls bears testimony to their necessity.” That is, if a teacher bothers to maintain a reward system, then that suggests to teachers and students that without the system, students wouldn’t have a reason to do the thing they’re being rewarded to do. I think many of us relate when we think about our hobbies and work. As Alfie Kohn puts it:
A lot of people have had the experience of having done something just because they loved it — until they started to get paid for doing it, after which they wouldn’t dream of doing it again without getting paid. The phenomenon whereby extrinsic motivators cause intrinsic motivation to evaporate is not on the tips of our tongues, but it’s not that far from consciousness, either.
The obvious objection is that there are activities in which students have essentially zero intrinsic interest. For these cases (which unfortunately seem prevalent), Lepper and Greene suggest that using rewards is “necessary and appropriate,” but Kohn disagrees. For what it’s worth, even a quick search of “ways to get children to do chores” yields plenty of options that don’t involve explicit rewards, and perhaps we can apply those ideas in classrooms.
There were actually six groups: high/low/no surveillance and expected/unexpected reward, but the experimenters found no differences between the high/low surveillance groups so they collapsed those two variables into one.
On footnote 2, the linked video has been taken down.