The Common Sense Everyone Follows
Every culture on earth celebrates intelligence. Parents beam when report cards come home and say, "I always knew you were a smart kid." Teachers single out gifted students. Self-esteem campaigns reinforce the message: let children know how bright they are, and you fuel their confidence to tackle hard things. The logic seems unassailable. If a child knows she's smart, she'll believe in herself when things get tough.
The Experiment That Broke the Logic
In 1998, Claudia Mueller and Carol Dweck at Columbia University ran six studies on 412 fifth graders to test that assumption directly. The design was simple. In the first round, each child solved ten pattern-recognition problems from the Raven's Progressive Matrices, a well-known intelligence test for children. The problems were chosen to be easy enough that nearly everyone scored well.
Then came the sentence that would change everything. Children were randomly assigned to one of three groups. The first heard: "That's a really high score. You must be smart at these problems." The second heard: "That's a really high score. You must have worked hard at these problems." The third, the control, heard only: "That's a really high score."
Five seconds of praise. One sentence.
In round two, every child received a set of problems designed to be too hard for their age group. Everyone failed. The researchers told each child their score was much worse.
Round three brought problems just as easy as round one. The question was simple: after identical success and identical failure, what did one sentence of praise do to their performance?
The Split
Children praised for effort solved about one more problem in round three than they had in round one. They treated the failure as a signal to try harder. When offered a choice between harder and easier problems next, roughly 90 percent picked the challenging set. They reported enjoying the tasks.
Children praised for intelligence solved about one fewer problem than they had in round one. Failure didn't motivate them; it deflated them. When offered the same choice, a majority picked the easy problems. They reported less enjoyment. Many showed visible frustration during round two, where effort-praised children described the same hard problems as their favorites.
The most unsettling finding had nothing to do with puzzles. In one experiment, children were asked to write a letter to students at another school describing the problems and their own score. Children praised for intelligence were significantly more likely to inflate their results. Nearly 40 percent misrepresented their performance. Among effort-praised children, the rate was far lower.
Why One Sentence Does This
Mueller and Dweck's explanation is structural, not moral. Telling a child she's smart frames intelligence as a fixed trait, something she either has or doesn't. Succeeding confirms it; failing threatens it. The rational move for a child protecting a smart identity is to avoid situations where failure is possible, quit when failure arrives, and conceal evidence that it happened.
Telling a child she worked hard frames effort as the cause. Success means the strategy worked; failure means the strategy needs adjustment. Nothing about the child's identity is on the line, so harder problems become interesting rather than threatening.
The distinction sounds subtle in the abstract. In data, it produces a two-point swing on a ten-point scale and a reversal in task choice, persistence, emotional response, and honesty. All from a single spoken sentence after a single success.
The Strongest Counterargument
Yue Li and Timothy Bates ran a series of replication studies in 2019 with more than 600 Chinese fifth graders and found largely null results. The praise manipulation did not reliably produce the performance differences Mueller and Dweck reported. Bates has been a persistent critic of growth mindset research, arguing that the original effects are too fragile to survive methodological scrutiny outside the original lab.
The rebuttal matters, but so does its context. Dweck and David Yeager published a response showing that Li and Bates's second and third studies had reduced statistical power below standard thresholds for replication, and that when all three datasets were analyzed together, the core effect reappeared. More critically, Li and Bates did not create a genuine failure experience: their children scored nearly 2.5 standard deviations higher on the difficult problems than Mueller and Dweck's original participants did. If a child doesn't actually feel like they failed, the mechanism has nothing to act on.
Separately, Gunderson and colleagues videotaped 53 families with toddlers (ages 1 to 3) during natural interactions at home and returned five years later. Children whose parents had used more process praise at ages 1 to 3 preferred challenging tasks and endorsed the belief that traits are improvable at ages 7 to 8. A 2018 Frontiers in Psychology study of Chinese schoolchildren found that ability-praised children engaged in more self-handicapping after failure, extending the phenomenon into self-worth protection theory.
The debate is real and not fully settled. But it is a debate about boundary conditions and effect size, not about whether the phenomenon exists.
What We Didn't Prove
The original study used pattern-recognition puzzles in a controlled lab setting with predominantly American fifth graders. Whether the effect generalizes equally across cultures, ages, and real classroom environments remains an open question. The Li and Bates replication in China found weaker effects, and the mechanism may operate differently in collectivist cultures with different relationships to effort and intelligence.
All experimental studies measured short-term responses over minutes to hours. Gunderson's longitudinal work shows associations extending over five years, but it is correlational and cannot establish that praise type alone causes lasting mindset changes.
No study has tested whether occasional intelligence praise embedded in a pattern of effort praise produces harm. The research contrasts extremes. Most parents naturally mix both types, and the consequences of an occasional "you're so smart" within a broadly effort-oriented approach remain unstudied.
The Bottom Line
Telling children they're smart after they succeed feels generous. Six experiments showed it backfires. The child who hears "you're smart" learns that intelligence is the explanation for success, which means failure becomes an indictment of who they are. The child who hears "you worked hard" learns that effort is the explanation, which means failure becomes a problem to solve. One sentence, delivered once, after one success, was enough to flip whether children chose challenges or avoided them, persisted or quit, told the truth or lied. The finding is 28 years old, has survived reasonable scrutiny, and costs nothing to act on.
What You Can Do
When a child succeeds, describe the process, not the person. "You stuck with that even when it got confusing" works. "You tried a different approach on the second problem" works. "You're so smart" does not, at least not if you want the child to stay curious when things get hard. This applies to teachers, coaches, grandparents, and anyone who talks to children about their performance.
When a child fails, resist the urge to console with "you're still smart." The child who believes smartness is fixed hears that as denial of evidence, not comfort. Instead, treat the failure as information: what didn't work, and what could change?
The shift is small, habitual, and free. It requires retraining a reflex, not buying a curriculum.