Grit is hard to define and obviously desirable. Both halves of that sentence are true, and the interesting thing is that they are connected. The reason the word resists definition is the same reason it keeps producing mixed results, and the research on it now makes that unusually clear.
What grit was built to predict, and did
It is worth starting with the case for it, because the criticism only makes sense against what came first.
Angela Duckworth and Patrick Quinn defined grit as trait-level perseverance and passion for long-term goals, built from two components: consistency of interest, meaning you do not keep switching what you care about, and perseverance of effort, meaning you keep working at it. Their validation of the short scale appeared in the Journal of Personality Assessment.
The Artful Age
A weekly letter on aging well, family across generations, and the creative life after the kids leave home.
The West Point result is the one that made the idea famous. Among 1,248 cadets, a score one standard deviation higher on the grit scale corresponded to 1.99 times the odds of completing summer training, after the modeled covariates. That is 99 percent higher odds, not a 99-percentage-point increase in probability. Because completion was already common, the distinction matters. Grit also added predictive information after West Point’s Whole Candidate Score, which itself was not significant in that model.
A second study followed 190 National Spelling Bee finalists. A one-standard-deviation higher grit score corresponded to 1.38 times the odds of advancing to a later round. The authors’ mediation analysis linked grit, self-reported cumulative practice and performance, but the observational design did not establish that grit caused more practice or that practice carried a causal effect.
What happened when the whole literature was pooled
Nine years later, Marcus Credé, Michael Tynan and Peter Harms gathered 88 independent samples, 584 effect sizes and 66,807 people, and published the synthesis in the Journal of Personality and Social Psychology.
Two numbers do most of the damage. Grit correlated with conscientiousness at .84, a meta-analytic figure corrected for measurement error, and high enough that the two are difficult to tell apart as constructs. And once conscientiousness was accounted for, grit explained essentially no additional variance in academic performance, with a change in R squared of .004 or less.
Grit’s own correlation with academic performance came out at .18, and with retention between .12 and .18. Those are real relationships. They are also considerably smaller than the word’s reputation implies.
The part that explains the definition problem
Here is the finding that makes sense of the whole thing.
The two halves of grit did not behave like two halves of one trait. Perseverance of effort correlated with academic performance at .26. Consistency of interest managed .10. Against conscientiousness, perseverance came in at .83 and consistency at .61. The authors could not confirm the higher-order structure at all, and concluded that the two components function as separate constructs rather than as facets of a single thing.
That is why the word is hard to define. It is not vague in the ordinary way that character words are vague. It is a compound of two dispositions that the data keeps pulling apart, and most of the predictive weight sits on one of them.
What survives, and it is not nothing
Read carelessly, that reads as a debunking. It is not, and one number in the meta-analysis makes the point.
Using hierarchical regressions built from the meta-analytic correlation matrix, perseverance added estimated R-squared changes of .040 for overall academic performance, .085 for high-school GPA and .023 for college GPA after conscientiousness. These are synthetic incremental-validity estimates, not results from one direct test or an intervention. The authors also noted that such increments could appear even if perseverance is another manifestation of conscientiousness measured by a different inventory.
The cautious conclusion is that the perseverance facet carried more of the academic association than consistency of interest, while the full grit score contributed little beyond conscientiousness. Whether perseverance is a distinct causal quality that can be strengthened remains unresolved.
What neither paper settles
Both sets of authors are candid, and it would be unfair to quote one side’s caveats and not the other’s.
Credé and colleagues note that most of the studies they pooled used concurrent designs, which can inflate the relationship between grit and performance, and that self-reports of a socially desirable trait carry the usual problems. Retention in military and academic settings has a high base rate, which mechanically dampens correlations. They also lacked the power for full moderator analyses.
Duckworth and Quinn note the self-report problem themselves and addressed part of it by validating informant versions, so that other people rated a participant’s grit. They also flag, before anyone else did, that their conscientiousness measure was a brief inventory rather than the full instrument, and that whether grit differs from conscientiousness’s achievement-striving facet needed more work. It is worth saying that Duckworth is the author of the construct being examined here, and that she raised this particular question first.
Where that leaves the word
Probably where it started, but with better instructions.
The meta-analysis supports an association between the perseverance facet and academic performance. It does not show that teaching perseverance raises performance, or that consistency of interest never matters in other settings.
The evidence supports scoring the two facets separately and treating grit as a contested higher-order construct. It does not settle where perseverance ends and conscientiousness begins.