There’s a well-known finding in intelligence research called the “worst performance rule” (WPR): when people do a reaction-time task, their slowest responses predict IQ better than their fastest ones. The idea has been intriguing to researchers because it suggests something counterintuitive. Whatever gives rise to general intelligence (the “g factor”) might not be about how well your brain can perform at its peak, but about how badly it fails under strain. A new meta-analysis pooling 19 studies and nearly 3,800 participants set out to test just how solid this effect really is.
The short answer: the WPR is real, but it doesn’t work quite the way people assumed. Correlations between reaction time and intelligence did get stronger as you move from fastest to slowest responses (from about -0.18 to -0.33), but the increase wasn’t a straight line, it was logarithmic. Most of the jump in predictive power happened going from your fastest responses to your average ones; going from average to your very slowest responses added only a little more. In other words, it’s not so much that your worst moments are uniquely revealing, it’s that your best moments are unusually uninformative. The authors half-jokingly suggest a rename: the “not-best performance rule.” The analysis also found no support for some classic predictions, like the idea that this effect should get stronger in harder tasks or in less intelligent samples.
What does this mean for theories of intelligence? It’s a bit of a blow to accounts that treat your slowest responses as a special window into “lapses of attention,” since those predict a much sharper uptick at the slow end than what actually showed up. It fits more comfortably with drift-diffusion-style models of decision-making, though the authors note those models are flexible enough to fit almost any pattern, which makes them harder to use as a real test. The bigger takeaway is a note of caution: this tidy, often-cited effect turns out to be less clean and less central to explaining what makes people “generally intelligent” than the existing literature suggested.
Link to full article: Redirecting