A Japanese rising negative question accompanies an implicature that the speaker has a bias toward the positive answer. The rising intonation for the construction has two variants: one with a lexical accent of the predicate retained and one with deaccentuation of the predicate. Introspection-based data suggests that the speaker uses deaccentuation when the speaker has public evidence stronger than hearsay or circumstantial evidence. To confirm the correlation between deaccentuation and availability of public evidence, we conducted a naturalness rating experiment to test the hypothesis that evidentiality affects the choice of the intonational patterns. The experiment shows that deaccentuation marks an embedded proposition in a biased question as evidenced. The utterance results in a meta-question about the discourse states: the speaker asks whether the available evidence is clear enough to support the positive answer.