@Cas:

It’s one of the simplest ways to demonstrate this in conversation, as one’s interlocuter pauses to scratch their heads in confusion as they try to think of a single film that passes.

Hang on. Surely the purpose of this test is to actually demonstrate that women are underrepresented in films, not to demonstrate that the person you’re talking to can’t think of any off the top of their head? (Although personally I think there would be more value in demonstrating that women are inaccurately represented in films, which I suspect is rather more of a problem).

But to me, the aim of any test of societal norms should be to make a valid point, not just to get people to scratch their heads because they don’t have instant recall for the contents of film scripts. Surely you shouldn’t resist increasing the accuracy of the test out of concern that someone could then name one or two examples of films that pass: occasionally saying ‘Really? Just those two? Out of all the films ever made?’ is hardly going to sink your entire position, is it?

For example:
Under both the original and revised tests, Kenneth Brannagh’s Henry V would pass with flying colours. So would Titanic. So would Pretty Woman.

Does that mean that those three films are reasonable and accurate portrayals of women and their interests? Hell no. Does it mean that the Bechdel test might need to take a more detailed approach to be accurate? I’d say yes.

Greater granularity might make the Bechdel test a bit less fun to roll out mid conversation, but it’d help to bolster the quality of the results. (But, again, that’s me worrying rather more about the accuracy of the representation of women, which this isn’t a test for. Although I still think a test for that would be more useful. A damn sight harder to implement, mind, but a lot more useful.)