The Bechdel test is stupid

Note, the header image was used non-ironically on another website.

Tl;dr:  the Bechdel test comes up in discussions about women’s roles.  It is a useless test.  It neither explains the role of women in the novel nor does anything to make the novel better, hence my ironic usage of the header image.

For those unfamiliar with it, a Bechdel test consists of two parts, with a third part sometimes added:

1. Two women must talk to each other.
2. They must not talk about men.
3. Sometimes, it’s added that both women must be named.

It’s a very useless test.  Let’s make a hypothetical story that illustrates why.  A man and a woman are on a plane, when it suddenly has engine failure and goes down.  All of the passengers are dead, except the man and the woman.

The man is completely paralyzed.  The woman has to get all of the food, water, and build the shelter.  Part of the wreckage she recovers has a map on it.  She doesn’t know exactly where they are, so she builds a makeshift raft and goes from island to island, marking off places that she knows are not correct.  She can’t venture too far out though, because otherwise the man will die or at least be in serious trouble when she gets back.

After doing this for a few years, she figures out where they are.  The nearest civilization is a week away on a raft.  Now she has a choice.  If she chooses to leave the island alone, the man will die.  That’s been her only friend for years.  If she chooses to leave the island together, they might both die, since the man can’t do anything to help out.  Does she abandon the island?

That’s the story.  The woman is clearly the main protagonist, has all of the action, and basically does everything.  This hypothetical story would still fail the Bechdel test.  We don’t have two named women talking to each other in the story.

This is the exact scenario of one movie, Gravity, even though Sandra Bullock portrays the heroine, it fails the Bechdel test.

Part of any story is that there’s a main character.  Everything in the story will revolve around the main character.  If it doesn’t, you’ll have a meandering story with needless filler.  Needless filler is a bad thing.

So if you have a typical trifecta in a story, male protagonist, female love interest, and male best friend, the dialogue is going to revolve mostly around the male protagonist.  Most everyone will be talking about or reacting to what the male protagonist does.  This is basic storytelling.

So if you look at a list of movies that pass the Bechdel test, they all have one thing in common.  The unambiguous winners are  Beauty and the Beast, Star Wars: Last Jedi, Wonder Woman, Despicable Me 3, and Guardians of the Galaxy.  Of those, the first three are female protagonist lead. The last two feature scenes involving female family members.

The ones that rank on the barely scale are Spiderman Homecoming, It, Logan, and The Fate of the Furious.  These are male protagonist stories.

What you learn at a quick glance is that the Bechdel test is heavily weighted against any stories involving a male protagonist.  Since the story revolves around the protagonist, the only way around it with a male protagonist is to have female family members, or to insert random dialogue to get the barely mark ticked off.

Goodhart’s law states that:

When a measure becomes a target, it ceases to be a good measure.

When films actively try to tick off a mark to pass with barely, that mark becomes meaningless.

The Bechdel test is a convenient way to discredit any stories with a male protagonist without having to explicitly say, “I don’t like stories with male protagonists.”  It’s fair enough if you don’t, just don’t make up a test about it.

It’s a compelling theory because on the surface, it looks deep.  How hard is it to have two female characters talk about something besides a man?  But once you scratch it, you realize that it doesn’t tell you anything useful.

What would a decent test look like? Well, you need to ask what a test is even supposed to do.

There are three parts to any test. Construct-validity, does the test measure the things it is purporting to measure? Content-validity, does the test bias itself towards some groups more than others? And predictive-validity, does it actually predict anything useful?

In terms of content-validity, I say that it has none, and show that overwhelmingly, it favors certain types of content over others.

In terms of predictive-validity, it doesn’t tell me anything useful. Anything useful is derived from its content-validity, which is whether or not the movie is female protagonist.

http://www.telegraph.co.uk/culture/film/10450463/Bechdel-test-is-damaging-to-the-way-we-think-about-film.html

Movies that pass the Bechdel test include “empowering films like Showgirls… The Hottie and the Nottie, Bride Wars, Bratz: The Movie, Burlesque, Sex and the City 2 and any number of other stereotype-peddling car wrecks in which two or more of the leads happen to be female.”

That’s a low level of predictive power there. As stated, that’s derived from its content-validity, it’s simply a measure of two leads who are female.

For Bechdel it was never a test, but the trigger for a pessimistic punchline: even in films that meet these seemingly foolproof criteria, the female characters are always delimited by something.

As for how predictive it is:

three of those five results are either false positives or false negatives: for that level of insight into culturally ingrained sexism, you might as well forget the Bechdel test and go straight to Jim Davidson.

In short, it has no construct validity and therefore, no predictive validity. It is useless as a test, from all definitions we have of what a test is supposed to do.

So, what would a good test look like? It wouldn’t be anything as easy or shallow as the Bechdel test.  There’s been a number of alternatives proposed to the Bechdel test, most of them are still as shallow as the Bechdel test.

Each of these alternative proposals are simple bean counting exercises, the only question is what color beans do we count.  When dealing with a complex phenomena, attempts to reduce that phenomena to a single point are necessarily inaccurate and misleading.  So here’s a complex list.

1.) Agency. Do the females have a role in the book besides “wife or hot girlfriend” of protagonist?  Do they exist for reasons outside of serving the interests of the protagonists like giving them emotional pathos or story resolution?  Or being rescued ad nauseam like Lois Lane in a Zack Snyder film?

Note that female protagonist roles would fail this as well.  The exemplar here is Lara Croft’s rebooted Tomb Raider (2013) series.  The only reason that side characters exist is so Lara can have whatever emotional development she needs at that point before the characters are immediately discarded.

Sound like other criticism I’ve leveled at books?  Yes, this is rule 5 of the sins, NPCs are not replaceable, or the girl in the fridge problem.  Side characters should have their own behaviors and motivations outside the sphere of the main character because it’s a principle of telling a good story.  This applies to both male and female characters.

2.) Pivotal role. Do the females have a major part to play in the book? Could you remove them?  Again, this is the same criticism I have when discussing good and bad harems.  A harem without interesting people in it is boring.

Our test is pretty simple:  A work should have women with agency doing things that actively change the story.  You can’t count beans with that theory, but it’s a better formula than the rest of what’s out there.  If you do this instead of worrying about how many women talk to each other about nonsense, you’ll get the added bonus of having a better story than following all the bean counter versions.

3.) Realism.  Now we have the exception that requires more analysis than anything as simple as the Bechdel test would ever allow.

Let’s say I write a book about Athenian Greece. In Athens, women weren’t allowed outside without their husband and were basically confined to domestic duties. It would make complete sense if women didn’t appear much in a novel that revolved around ancient Athens.

Now let’s say I do one in Spartan Greece. The men were constantly out in battle, Sparta was run by women. It would not make sense to do a novel about Sparta and not mention all the women on it, at least if the novel is about Sparta itself and not about the Spartan men in war.  If it’s Spartan men at war, there weren’t any women.  Realism depends upon the context.

This applies to other contexts too.  Doing a game about WWII?  Sorry, there weren’t any female infantry soldiers in the Allied troops.  Doing a medieval story about the Holy Roman Empire?  You won’t have any racial diversity in it.