Do readers with weakly held priors respond more to linguistic features than ideologically committed ones?

This reads the question as asking whether linguistic craft matters most for readers whose convictions are loose, and matters little for the ideologically locked-in. The corpus has one note that speaks almost directly to the underlying mechanism — and it actually inverts the usual assumption. Analysis of debate corpora found that a reader's political and religious ideology predicts whether they're persuaded *more* than any feature of the language used Does what readers believe matter more than what debaters say?. The striking twist: language effects that *appear* in studies without controlling for who's in the audience turn out to be confounds — what looked like persuasive phrasing was really audiences self-sorted by topic. So before assuming weakly-held readers are linguistically movable, the corpus warns that much of the apparent power of wording is an artifact of who showed up to be measured.

That said, the corpus does suggest a place where weak priors leave a reader exposed: not to argument, but to framing. When a message smuggles in a false assumption rather than asserting it outright, even systems that *know* the correct fact tend to accept the buried premise — false presuppositions drive far more accommodation than correct knowledge drives rejection Why do language models accept false assumptions they know are wrong?. The lesson generalizes: a reader without a firm prior has nothing to push back against a presupposition with, so the linguistic feature that matters isn't eloquence — it's what gets quietly assumed.

There's also a subtler channel the corpus surfaces: persona. A large study of writers and readers found that surface linguistic assistance systematically shifted how readers perceived the *writer* — toward more confidence, more extremity, more authority — across every dimension measured Does AI writing assistance change how readers perceive the writer?. For an uncommitted reader, perceived confidence and authority of the speaker is exactly the kind of cue that substitutes for a prior. So linguistic features may move weak-prior readers less through the content of the argument and more through the *impression of the arguer* they create.

Finally, the corpus complicates the premise that readers share a stable response at all. Interpretation of socially-loaded sentences is irreducibly multiple — disagreement reflects valid differences in reader position, not noise Why do readers interpret the same sentence so differently?. That means "linguistic feature" and "prior" aren't cleanly separable inputs: the same wording lands differently depending on where the reader stands, which is itself a kind of prior. The honest synthesis: the corpus leans toward priors dominating language, treats much measured language-effect as confounded, and points to presupposition and persona — not persuasive phrasing — as the routes most likely to reach a reader who hasn't made up their mind.