Pronouns and Gender

How did you come by this topic?

My research concerning the linguistic encoding of gender was originally inspired by my work on the real-time comprehension of (locally) ambiguous sentences. For instance, in Frazier et al. (2015), we test sentences like (1), where the underlined words corefer with each other:

  1. a) Which cowgirl did Annie expect to have injured herself?
    b) Which cowgirl did David expect to have injured herself?

Both of these sentences are grammatically licit, but both also contain substrings like in (2), which have distinctly different coreference properties.

  1. a) Did Annie expect to have injured herself?
    b) Did David expect to have injured herself?

It’s possible that the parser, upon reaching herself in (1), looks leftward to figure out who herself refers to. As it looks back through the sentence, and sees a string that looks a lot like (2), so it tries to link herself and the name (Annie or David). If it tries to link with the name in these sentences, we should find sentences like (2b) have a slowdown in reading times as compared to (2a) because linking herself to David shouldn’t work.

Because, we assume, herself has the property of and David has the property of . But… this assumption is a lot more complicated than most people realize!

For instance, (3) is probably a completely licit and acceptable sentence, but only because of what we know about Halloween parties:

  1. At the Halloween party, the cowgirl accidentally dropped his lasso in the kitchen.

Gender of names

David is a pretty common name for men and boys, so that might be obvious. But what about a name like Taylor? Taylor is one of the most balanced names, in that in both the UK and USA, there are surprisingly similar numbers of men and women named Taylor. Does Taylor have a property like or ? Is it underspecified? Or does it have both properties?

What about a name like Michael?

Michael is a very common name for men and boys… but if you’ve been keeping up with Star Trek: Discovery, the main character is named Michael, but she’s a ciswoman. The characters to occasionally comment on the unusualness of a woman with a masculine name, but otherwise adjust easily and never use the wrong pronouns for her. I suspect this is how it would play out in real life, too. So if we can adjust the proper pronouns to corefer with a name so easily, how can we assume David and herself mismatch in some gender property?

Very importantly: the gender of the person who has a given name is independent of the gender bias associated with the name

Gender of pronouns

This is particularly relevant for people whose gender is nonbinary, i.e., are neither women nor men. If someone is nonbinary and isn’t “a she” or “a he”, then it doesn’t matter what their name is, the standard pronouns like she/her/herself and he/him/himself are wrong and inappropriate. Moreover, using the wrong pronoun for someone (i.e., misgendering them), has been shown to cause measurable and lasting damage to them.

So what’s the property of human gender that pronouns encode?

Screen Shot 2017-10-20 at 1.46.37 PM

Selected works:

Also, check out this article at The Economist that references my work!