The Flynn effect is the tendency for IQ scores to increase over time. It is understood that some subtests or tasks show a stronger Flynn effect than others. But what about specific test questions?
A new study investigates the Flynn effect on individual math test items. From 1986 to 2004, the researchers found that some items showed a consistent increase in passing rates. Sometimes passing rates increased by 10 percentage points (or more)!
On the other hand, other items showed no change or even a drop in passing rates, an “anti-Flynn effect.” The authors also tried to identify characteristics that differed across FE, anti-FE, and other items.
The result was that Flynn Effect items were usually story problems about real-world applications of math. Here are two examples of the type of items that show a positive Flynn Effect in the study. (Note: these aren’t real items from the test; those are confidential.)
Items showing an anti-Flynn Effect measure learned knowledge or algorithms for solving problems. In other words, there is no real world application; these items just measure whether a child has learned information explicitly taught in math classes.
The lesson is clear: in the late 20th century, American children got better at solving math problems that were presented in ways that required applying math to solve real-world problems. But children became less adept at using formulas and math knowledge to solve abstract questions.
It’s a fascinating study that gives a hint about why certain tests show Flynn effects and others don’t.
This study completely validates the current debate in education. We’ve spent decades prioritizing “story problems” and “real-world application” (the FE items) over explicit, structured instruction in algebraic algorithms and geometry theorems (the Anti-FE items). It seems we’ve successfully trained kids to apply math heuristically, but they are losing the ability to manipulate abstract symbols and formulas. Great for taxes, bad for theoretical physics.
@noah_gowie342 I agree. We’re effectively optimizing the K-12 math pipeline for careers that require basic financial literacy and data interpretation, but we are inadvertently hindering the development of the next generation of engineers and pure mathematicians who absolutely must be fluent in symbolic abstraction. This has real national competitiveness implications.
The item-level analysis reveals something crucial - Flynn gains aren’t uniform cognitive improvements but specific to certain problem types. Real-world reasoning items (money/banana problems) showed FE gains while algorithmic items (triangle definition, algebra formulas) showed anti-FE declines. This suggests cultural/educational shifts toward applied problem-solving at the expense of procedural knowledge. It explains why fluid reasoning tests show strong FE but crystallized knowledge tests don’t - the effect is domain and task-specific, not general intelligence increase.
This pattern matches the shift from rote computation to conceptual understanding in U.S. math education since the 1980s. Students improved at flexible reasoning but lost automaticity with procedures. The anti-FE on geometric definitions and algebraic manipulation is concerning - these aren’t outdated skills, they’re foundational. The question is whether trading procedural fluency for applied reasoning is optimal, or if we need both. STEM fields require both conceptual flexibility AND computational mastery.
I wonder if the anti-Flynn on procedures partly reflects that calculators/computers made some of these skills less critical for daily life. That said, you’re right that STEM fields still need them. But maybe the question is when to build procedural fluency? Perhaps conceptual understanding first, then procedures, works better than the reverse? Though the data here suggests we just… stopped teaching the procedures effectively.
That’s an interesting dilemma. I’m curious whether this trade-off is inevitable or if it’s specific to how reform math was implemented in the U.S. It does seem like losing foundational skills undermines the goal of better problem-solving in the long run.
