An effect being significant in one sex but not the other is NOT evidence of a sex difference

https://xcancel.com/NeuroSGS/status/2087959408122700224

There’s a surprisingly common statistical mistake when researchers investigate sex differences in the brain and behaviour, and this new study (below) reveals just how bad this problem is.

Imagine hippocampal volume is significantly associated with memory performance in males (p<0.05), but NOT in females (p>0.05).

Does that mean the relationship differs by sex?

No.

“Significant in males” + “not significant in females” isn’t, by itself, a significant sex difference.

Instead, we need to test whether the association itself differs by sex. For example:

Memory ~ Hippocampal volume × Sex + covariates.

The Hippocampal volume × Sex interaction tests whether the relationship between hippocampal volume and memory differs between males & females.

Surprisingly, this statistical misunderstanding is surprisingly common in the literature.

A new PNAS (2026) study examined 200 papers whose titles claimed sex/gender-dependent effects in the behavioural and brain sciences.

Only 24% actually provided appropriate statistical evidence for the claim that was tested.

58% never statistically compared the effects between sexes (the interaction effect).

Unfortunately for me, in the neurosciences, only 18% of claims were appropriately supported statistically.

Overall, an effect being significant in one sex but not the other is NOT evidence of a sex difference.

https://www.pnas.org/doi/10.1073/pnas.2608703123