[zmgsabst]

Yes — discussing a bimodal distribution as a bimodal distribution is useful, even in the presence of outliers and data points bridging the two peaks.

“Man” and “woman” are names for those nodes in the bimodal distribution of traits, as correlated with sex. Same as “cow” and “bull”, or “hen” and “cock”, or “doe” and “stag”, or “female” and “male”.

I’d rather my language be able to discuss the experience of the 99%+ than become incapable of discussing basic facts (like apes being sexually dimorphic) because reality might offend outliers.