The sound of accurate recognition memory decisions

Metacognitive confidence in memory judgments is typically assessed with a numeric self-report rating scale, a measurement approach that reliably predicts judgment accuracy but may also capture individual differences unrelated to memory per se. Recent research in perceptual discrimination suggests that the acoustical features of verbally rendered cognitive judgments (i.e., prosody) may provide relatively automatic and direct cues to the accuracy of those judgments. The current study tested whether prosody would predict the accuracy of spoken long-term episodic memory judgments. Subjects studied and were tested on memory for faces in a forced-choice recognition procedure. Test responses were given by saying “Number One/Two/Three/Four” to indicate the selected face. The pitch, loudness, speech rate, and onset time of these responses were extracted and used as predictors of accuracy. Despite a retention interval in the tens of minutes and the brief, generic nature of the verbal utterance, all four speech signals discriminated accurate and inaccurate responses: Correct recognition judgments were higher pitched, louder, faster, and initiated earlier than incorrect judgments. The same pattern of results was observed comparing judgments on more difficult (4AFC) versus less difficult (2AFC) trials. Modeling analyses demonstrated that pitch and loudness provide redundant predictive information with speech rate and onset time, and that speech rate and onset time predict accuracy above and beyond explicit confidence ratings. Prosodic features thus appear to carry information about the accuracy of memory reports, and may indeed help humans make metamnemonic inferences of others.