The contempt you're bracing for

There is a well-documented gap between how much Americans think the other side despises them and how much the other side actually reports despising them. The political-science literature calls it the perception gap: ask partisans what the other team believes about them, and they guess far darker than the truth. People imagine a contempt that is, in large part, a projection.

This project set out to test that against language. If mutual contempt is overestimated, then a lot of what looks like contempt in text shouldn't be people aiming a slur at a target at all. It should be people reporting that others look down on them — "to them we're just flyover country" — or claiming the word with pride. We call that split the Attribution Asymmetry Index.

First, the data said we were wrong

The first real sample — about a thousand Reddit comments across twelve terms — did not show what the hypothesis predicted. Term by term, the contempt words were used far more to aim than to attribute.

Loading data…

redneck came back 83% aimed. yuppie, 95%. If you stopped here, the honest conclusion is that the debunk fails: most contempt words are genuinely thrown, not imagined. We almost wrote that essay.

Then we asked who was talking to whom

The averages were hiding the actual structure. Each of those per-term numbers pools two completely different conversations — and once you separate them by direction, the picture inverts.

Loading the map…

When contempt crosses between groups, it is overwhelmingly aimed. Liberals talking about conservatives: 93% direct. The coastal/rural axis: nearly all direct. People really do throw these words across the divide, and the targets are not imagining it.

But inside a group, the same words do the opposite work. Rural people talking among themselves about "rednecks" are reclaiming it — 80% claim, worn with pride. Conservatives talking among themselves about "deplorables" and "bitter clingers" are attributing — 74% of it is "this is what they call us." The imagined, attributed contempt is real. It's just a conversation a group has with itself about how it's seen — not the contempt actually flowing the other way.

Same word, different room

It goes further than direction. The same contempt word does opposite work in different communities. Take "redneck": in political subreddits it's thrown as an insult; in regional and identity subreddits it's worn with pride.

Loading…

Read across the rooms and the bars flip from mostly-aimed to mostly-claimed. The word didn't change — the room did. Whatever "redneck" means is less a property of the word than of who's saying it and to whom. That's the same lesson as the direction split, sharpened: contempt isn't a fixed quantity attached to a slur, it's a thing people do with words, and they do very different things depending on which room they're standing in.

The tell is in the word

There's a second pattern. The terms that lean attributive are precisely the ones that were coined as accusations — usually from a famous quote. Nobody earnestly calls someone a "bitter clinger" as their own judgment; they say "they called us bitter clingers." So that term is 64% attributed. Ordinary epithets you'd actually call someone — yuppie, white_trash — are aimed. The grammar of the word predicts how it travels.

What this is, and isn't

A few honest limits. The classifier is not yet gold-validated — the attribute-versus-aim call is the hardest one, and the numbers above are suggestive, not settled. The in-group edges lean on speaker inference, which is noisier than target inference. And term-matching can only see contempt that contains the word — attributed contempt phrased without the slur ("they think we're stupid") is invisible to us, which likely undercounts attribution further.

But the shape is interesting enough to state plainly: the contempt you brace for is mostly something your own side says about how you're seen. What actually crosses the divide is aimed — and rarer than the bracing suggests. That's not the tidy debunk we set out to prove. It's a better one.

Take the perception-gap test and see your own gap — then come back and read the receipts on the map.