+At this point in my ideological evolution, I _did_ have a concept of Bayesian reasoning as normative ideal. But I thought to myself, well, base rates: _most_ people aren't gay, and the professor's voice isn't _enough_ evidence to overcome that prior; he's probably not gay.
+
+Looking back, there's nothing wrong with the _form_ of my reasoning—[base rate neglect](https://en.wikipedia.org/wiki/Base_rate_fallacy) is in fact a thing—but I suspect I was _quantitatively_ in the wrong? If 3% of men are gay, you "only" need log<sub>2</sub>(97/3) ≈ 5 bits of evidence to think that someone probably is. Is a sufficiently distinctive "gay voice" that much evidence—something you're 32 times more likely to hear from a gay man than a straight man?
+
+I ... actually think it plausibly is? I think I was reaching for "base-rate neglect" as an excuse for my old egalitarian prior that stereotypes are invalid. But even if the likelihood ratio isn't quite that large—if I only had 3 bits of evidence that L. or Prof. H. were gay, that's still an update from 3% to 20%. [Gaydar](https://en.wikipedia.org/wiki/Gaydar) is real! Of _course_ it's real (to some quantitative extent).
+
+[TODO: incorporate studies footnoted in https://en.wikipedia.org/wiki/Gaydar to see how big the voice effect actually is]