Effective human cooperation hinges on the ability to identify unreliable individuals. Understanding how this "epistemic vigilance" operates from naturalistic sensory inputs remains a challenge. While language provides explicit markers of certainty (e.g., "I don't know") and evidentiality, these are resource-intensive. This study explores the hypothesis that simpler, low-level mechanisms, such as prosodic cues, enable rapid, automatic detection of unreliability across languages and cultures. Previous research has linked uncertainty in speech to decreased volume, rising intonation, and slower speech rate, but the precise perceptual representations and the relationship between perceived certainty and honesty remain unclear. These studies often rely on elicitation procedures where speakers deliberately produce these cues, potentially confounding production and perception. Moreover, the co-variation of acoustic features complicates independent analysis of their influence on perception. The researchers suggest that the observed prosodic signatures may be natural signs of cognitive effort, potentially explaining their association with both uncertainty and dishonesty, as both are thought to increase cognitive load.

Existing research indicates listeners can infer speakers' certainty from pauses, fillers, gestures, and prosodic features. Studies typically involve two phases: encoders (actors or speakers) produce utterances with varying certainty levels, followed by listener judgments based on acoustic analysis. Findings show associations between uncertainty and decreased volume, rising intonation, and slower speech rate. However, these studies have limitations. First, the deliberate production of cues may not reflect natural communication. Second, the use of actors may introduce stereotypical rather than veridical expressions. Third, the co-variation of acoustic features limits the understanding of their individual contributions to perception. Finally, correlational designs don't fully elucidate the underlying mechanisms. The close resemblance between these prosodic signatures and those associated with cognitive effort (increased pitch variability, slower articulation rate) suggests they may be natural signs of cognitive effort rather than deliberate communicative signals. The study hypothesizes that a common prosodic signature underlying both perceptions of certainty and honesty might originate from physiological constraints on speech production related to cognitive effort.

The study employed four experiments with a total of 115 listeners (primarily French speakers, with cross-linguistic validation in English and Spanish). Study 1 used psychophysical reverse correlation to identify perceptual representations of honesty and certainty. Randomly manipulated pseudo-words were presented in pairs, and listeners chose which sounded more dishonest or certain, rating their confidence. Perceptual representations were derived from normalized temporal kernels, comparing acoustic features (pitch, loudness, duration) of chosen versus unchosen stimuli. Studies 2A and 2B acoustically manipulated speech stimuli to test the common prosodic signature in contextualized situations. Study 2A involved a Likert scale rating of certainty and dishonesty, with contextual information provided to influence listeners' interpretations. Study 2B investigated the relationship between perceptual judgments and explicit conceptual knowledge about epistemic prosody. Study 3 tested the cross-linguistic generalizability of the findings using English and Spanish speakers and a group of multilingual speakers with varying levels of French exposure. Study 4 examined the automatic impact of the prosodic signature on verbal working memory using an implicit memorization paradigm. Listeners memorized pseudo-words with manipulated prosodies and then performed a recognition task, allowing an assessment of how prosody influenced memory.

Study 1 revealed strikingly similar perceptual representations for honesty and certainty across pitch, loudness, and duration. Unreliable speech was characterized by rising intonation, less initial intensity, slower speech rate, and greater variability in pitch and speech rate. High agreement between certainty and honesty judgments across tasks, with higher confidence for consistent judgments, further supported the common signature. However, judgments of honesty were less stable and precise than certainty judgments. Study 2A confirmed the common signature in a new sample, but showed that providing contextual information about potential deceit introduced significant inter-individual variability in honesty judgments, unlike certainty judgments. Study 2B demonstrated weak links between perceptual and conceptual knowledge of epistemic prosody. Study 3 showed that the common prosodic signature was perceived across languages (French, English, Spanish), indicating language independence. Study 4 demonstrated that the prosodic signature of unreliability automatically improved verbal working memory, suggesting an attention-grabbing effect.

The findings strongly support the hypothesis that a common prosodic signature underlies listeners' perceptions of certainty and honesty. The similarity of perceptual representations across tasks, languages, and the independence from explicit conceptual knowledge suggest this signature is a natural sign of cognitive effort rather than a culturally learned convention. The automatic impact on working memory indicates an innate ability to detect unreliability. However, while certainty judgments largely reduce to perceptual decisions, honesty judgments involve more complex inferences and contextual interpretations, reflecting the potential for deceptive manipulation of prosodic cues. The differences between certainty and honesty judgments highlight the interplay between perceptual processes and higher-level cognitive inferences, with the context and listeners' biases playing crucial roles in interpreting prosodic displays.

This research reveals a core, language-independent prosodic signature indicative of unreliability in speech, impacting perceptions of both certainty and honesty. The automatic processing of this signature highlights an adaptive mechanism for detecting unreliability. Future research could explore the developmental origins of this mechanism, investigate the influence of tonal languages and non-default speech production patterns, and examine the interplay between prosody, phrasal structure, and semantic content in more detail.

The study primarily used pseudo-words, limiting the ecological validity. While previous research suggests findings generalize to real speech, further investigation is needed. The sample, while including cross-linguistic validation, was not fully representative of global linguistic diversity. Future research should address these limitations to further enhance the generalizability of these important findings.