Something did change in a famous set of American test results after 1990. Scores on several measures of creative thinking began to fall, even as IQ scores continued the long rise associated with the Flynn effect. The finding came from psychologist Kyung Hee Kim’s 2011 analysis of the Torrance Tests of Creative Thinking, drawing on 272,599 records collected across six rounds of national norming.

The youngest pupils often showed the clearest losses. Children from kindergarten through third grade had the largest fall in fluency, the ability to produce numerous relevant ideas, between 1990 and 2008. They also had the largest declines on several other subscales. That finding became the empirical heart of what Kim called America’s “creativity crisis.”

But the phrase is more confident than the evidence now allows. A 2021 critical reappraisal, using later test norms and more than 330,000 participants, found no evidence of a generational decline. The result worth explaining, then, is not a settled national diagnosis. It is a serious disagreement over how to measure change in something as broad as creativity.

What 272,599 actually counts

The headline number is enormous, but it can easily be misunderstood. Kim assembled summary data from the norming samples used for Torrance test editions in 1966, 1974, 1984, 1990, 1998 and 2008. The individual sample sizes were 3,150, 19,111, 37,814, 88,355, 54,151 and 70,018.

These were separate cross-sectional groups, not 272,599 children followed across their lives. The samples also included adults. The publisher described them as geographically balanced across the Central, Northeast, Southeast and Western United States, but reported neither its sampling procedure nor participants’ demographic characteristics. Researchers therefore cannot determine from the paper whether shifts in the composition of the samples contributed to the score changes.

This distinction does not erase a trend. Repeated norming samples are routinely used to detect population-level changes. It does limit what the trend can prove: no child’s creativity was observed declining through time, and “American children” is a broader population than the available test cohorts can securely represent.

What the Torrance test asks

Kim analysed the figural form of the Torrance Tests of Creative Thinking, or TTCT. Participants complete unfinished drawings, build pictures from simple shapes and add titles. Their responses are scored for fluency, originality, elaboration, abstractness of titles, resistance to premature closure and a group of creative strengths.

A comprehensive review of the TTCT describes a long-established instrument with evidence for reliability and predictive value, while also stressing that creativity is complex and difficult to compress into a score. Figural divergent thinking is best understood as creative potential under structured conditions. It is not a census of paintings made, inventions built, stories written or useful ideas carried into the world.

The distinction is familiar in modern creativity research. Generating multiple possibilities is one component of creative work, not its complete definition. Even walking’s measurable boost to divergent thinking, for example, tells us about idea production during a task rather than a person’s total creative life.

The declines were not all alike

For fluency, the one subscale scored consistently across the entire period, the largest fall from 1990 to 2008 appeared among kindergartners through third-graders. Its effect size was 0.32, conventionally described as moderate. The fall was smaller in grades four through six. High school scores did not decline significantly under Kim’s stringent threshold, while grades seven and eight and adults showed small, non-significant increases.

Originality offered a less dramatic picture. From 1990 to 2008, it declined slightly in the youngest band, with an effect size of 0.07, while it rose among seventh- and eighth-graders. Creative-strength ratings also declined, although age-specific comparisons were unavailable.

Elaboration produced the largest effects. From 1984 to 2008 it fell sharply across every age group, but adults had the biggest decline, at 1.54 standard deviations. Kindergarten through third grade came next at 1.23. From 1998 to 2008, abstract titles and resistance to premature closure fell most among the youngest group, though both effects were small.

The “biggest losses among the youngest children” description is therefore broadly accurate for fluency, titles and closure, and directionally true for originality. It is not true of every measure. There is another complication: scoring procedures changed in 1984 for every subscale except fluency. Kim accordingly split several historical comparisons into different periods. The prompts stayed the same, but the measurement ruler did not remain perfectly fixed.

IQ rose on a different track

The contrasting IQ trend is the Flynn effect, the long-observed tendency for later generations to score higher on many intelligence tests. A formal meta-analysis of IQ gains brought together 271 independent samples, nearly four million participants and data from 31 countries. It confirmed substantial gains across much of the twentieth century, though their size differed by ability, test and region.

Kim’s creativity and IQ figures were not two outcomes measured in the same children. Placing the trends side by side shows that test scores labeled intelligence and creative thinking could move differently. It does not show that rising IQ caused creativity to fall, or that either score captures every relevant mental ability.

The 2011 paper discussed possible explanations including standardized testing, screen use and reduced opportunities for unstructured play. Those were interpretations, not variables tested in the norming analysis. No school policy, technology or parenting practice can be identified as the cause from these data.

A reanalysis reached the opposite conclusion

Barbot and Said-Metwaly revisited the question with the 2017 TTCT norms added. They argued that the original claim depended on problematic decisions about comparisons, statistical methods, data presentation and interpretation. Their cross-temporal meta-analyses found neither a generational decline nor evidence for a creativity crisis.

Kim has not abandoned the original interpretation. In a 2021 update using norming data through 2017, she reported continued declines, particularly in imagination and among young children. In other words, adding more data did not produce agreement because the researchers also differed over how the data should be analysed and what conclusions the score patterns justified.

Evidence outside the norming series also resists a simple downward story. A meta-analysis of 41 studies, encompassing 40,918 children and adolescents, found an overall upward developmental trend in divergent thinking. It found no general fourth-grade slump, although results varied with the test, task domain, country and whether participants were identified as gifted. It did find evidence of a seventh-grade dip.

What can responsibly be said

Kim’s 2011 study identified meaningful declines in several TTCT scores, with especially clear changes among the youngest pupils on some measures. It remains important because it asked whether cognitive gains celebrated under one label might coexist with losses under another. Its unusually large dataset made the pattern difficult to dismiss as random noise.

Yet sample size cannot repair missing demographic detail, changing scoring rules or disagreement over the right statistical comparison. Nor can a divergent-thinking test alone establish that a generation is less creative in daily life. The finding is worth taking seriously, but it should not be read as the final word.

A stronger answer would combine stable measures across time with transparent, representative sampling. It would follow at least some children longitudinally and connect test performance to creative work beyond the testing room. Until then, the safest conclusion is narrower than “children stopped being creative”: one influential analysis found that several scores fell after 1990, while a later reanalysis of the testing record concluded that the national crisis itself was not there.