Yeah. The problem is that while syntax has rules, it is as much rhetorical as logical. This is why AI language is so much more difficult than AI math.GreenGoo wrote: ↑Tue Mar 26, 2019 5:05 pm edit: Doesn't matter. It's bugging me, that's all.
See, this annoys me. We already know that accusations aren't a felony, but that killing people is. We automatically know what is being referred to in the second sentence.
What if someone said this:
"They slandered me when they said I killed the mailman. That's a travesty. That's against the law".
What is the "that's" referring to in the second sentence? Or the third?
In your example, it's easy to assume that "that" refers to the slander simply because we expect that the speaker ("they slandered ME!") would feel that way.
It gets even harder if the example is this:
"They slandered Jane when they said she killed the mailman. That's a crime. That's against the law".
We can't with certainty know which way it tilts here without knowing how the speaker feels about Jane. There's no way to "do the grammatical math" and determine it logically. (But note that I changed your "travesty" to "crime" to make it even harder. Rhetorically, we're more likely to use "travesty" to refer to slander than to actual murder.)
The Sarah Sanders quote is hard for us because (a) we know Sanders worships Trump, and (b) it is no longer unthinkable for the Press Secretary to call POTUS critics traitors. Trump's destruction of norms has ruined context for us.
(Imagine the Sanders quote in the mouth of a Barack Obama Press Secretary: we wouldn't be having this conversation at all because it would be simply absurd to believe that an Obama official would suggest that critics were traitors.)
