All corrections
Substack July 27, 2026 at 10:18 PM

www.technotheoria.org/p/caffeine-works-in-some-people

1 correction found

1
Claim
aka 16 IQ points
Correction

The cited meta-analysis did not measure IQ, and its effect sizes are standardized mean differences from tasks like attention, reaction time, and driving performance—not IQ-point changes.

Full reasoning

This is an invalid conversion.

The cited paper is a meta-analysis of attention tests, executive function, reaction time, information-processing tasks, and driving measures after sleep loss—not IQ tests. Its reported Hedges’ g values are standardized mean differences (SMDs) for those outcomes.

Methodologically, an SMD is a unitless effect size. As the Cochrane Handbook explains, the standardized mean difference is used when studies measure outcomes in different ways, and the overall effect is reported in units of standard deviations rather than in the original units of any measurement scale. That means you cannot simply relabel an SMD from heterogeneous cognitive and driving tasks as “IQ points.”

So even if one computes some average standardized effect from the paper, calling it “16 IQ points” is incorrect because the study did not estimate IQ change at all.

2 sources
Model: OPENAI_GPT_5 Prompt: v1.16.0