Actual and perceived cues to deception: A closer look at speech

This 1982 study examined the relationship between actual verbal and paralinguistic cues associated with deception and the cues that people perceive as indicative of lying. Using 40 senders and 80 judges across five presentation modalities, the researchers found that perceived deception cues showed meaningful correspondence with actual deception indicators, though the relationship was imperfect.

Forty senders produced truthful, deceptive, and mixed affective statements about people they liked, disliked, felt ambivalent about, or felt indifferent toward. Eighty judges evaluated these messages across five distinct presentation conditions (tone of voice, visual, verbal transcript, undistorted audio, full audiovisual), enabling systematic isolation and analysis of specific verbal and paralinguistic features.

This foundational study investigated whether people's beliefs about deception cues align with empirical reality. Forty participants generated truthful and deceptive descriptions of individuals under controlled conditions, while eighty observers evaluated these messages across multiple communication channels. Researchers systematically coded verbal content (positivity, neutrality, self-references, specificity) and paralinguistic features (speech rate, filled pauses, nonfluencies). Results revealed that actual deception cues—including reduced positive language, increased neutral language, and higher nonfluency rates—significantly correlated with observers' intuitive judgments about deception, supporting the validity of folk understanding about verbal deception markers. However, observers also weighted certain cues (e.g., speech rate) more heavily than their empirical association with deception warranted, highlighting the imperfect alignment between intuition and reality.

This work established that naive perceivers possess genuine insight into verbal deception cues, validating foundational assumptions underlying behavioral analysis and interview-based deception detection methods. The findings underscore the importance of empirically validating intuitive deception judgments and identify specific paralinguistic features that warrant further investigation in both laboratory and field-based polygraph research.

Background & Context

This 1982 study emerged during a pivotal period in deception research when scientists were beginning to systematically examine the disconnect between relatively few studies of verbal cues to deceit despite evidence that speech could reveal deception. The research addressed a fundamental question that remains central to both polygraph science and behavioral analysis: Do the cues that actually reveal deception correspond to the cues that people believe signal lying?

The study was groundbreaking in its dual focus. The researchers examined specific verbal and paralinguistic cues that might reveal when deception is occurring or that might be used by perceivers in their attempts to detect deception; also, they examined quantitatively the correspondence between actual cues to deception and perceived cues to deception. This approach built upon Ekman and Friesen's earlier work on nonverbal leakage while extending the analysis specifically to speech patterns—an area that would later influence both traditional polygraph practice and modern statement analysis techniques.

Research Design & Methodology

Forty senders described people they liked and people they disliked ("pure" messages); and people they felt ambivalent about and people they felt indifferent toward (mixed messages). They also described the people they liked, pretending to dislike them and they described the people they disliked, pretending to like them (deceptive messages). Eighty subjects then made judgments about the senders under one of five conditions of presentation: tone of voice (content-filtered), visual, verbal (transcript), undistorted audio, and full audiovisual.

This experimental paradigm allowed researchers to isolate specific channels of communication. The study employed a sophisticated factorial design with multiple presentation modalities, enabling the team to determine which verbal and paralinguistic features actually correlated with deception versus which features observers used to make their judgments. The deception task involved affective misrepresentation—pretending to like or dislike individuals—which created motivation without the ethical concerns of high-stakes lies.

Key methodological features included:

• 40 senders producing truthful, deceptive, and mixed affective messages
• 80 judges evaluating messages across five distinct presentation channels
• Systematic coding of specific verbal content (positivity, neutrality, self-references, other-references, specificity)
• Analysis of paralinguistic features (speech rate, filled pauses like "um" and "er", nonfluencies)
• Quantitative correlation analysis comparing actual deception indicators with perceived cues

Results & Key Findings

For the cues that we studied, the degree to which the cues actually were associated with deception corresponded significantly to the degree to which perceivers used those cues as signs of deceit. This represented a crucial validation: laypeople's intuitions about verbal deception cues showed meaningful correspondence with reality, though the relationship was far from perfect.

Specific findings for actual deception cues included:

• When senders pretended to like people they really disliked, their descriptions were less positive and more neutral than when they honestly described people they really did like
• When feigning disliking, senders uttered more nonfluences than when expressing honest disliking
• Deceptive messages showed measurable differences in language tone and emotional valence
• Speech disfluencies increased during certain types of deceptive communication

Perceived deception cues revealed additional patterns. Perceivers judged as deceptive descriptions that were spoken slowly and contained many um's and er's. Expressions of liking that contained many other references, few self-references, and many nonspecific (undifferentiating) descriptors were also perceived to be deceptive. Notably, some of these perceived cues (like slow speech) were used by judges but did not actually correlate with deception, highlighting the gap between stereotypes and reality.

Discussion & Significance

The study's most important contribution was demonstrating that actual and perceived cues showed significant but imperfect alignment. This finding has profound implications for polygraph practice and lie detection training. It suggests that human intuitions about deception contain a kernel of truth but are also contaminated by stereotypes—particularly the belief that nervous-sounding speech (slow rate, many hesitations) always signals dishonesty.

The research also advanced theoretical understanding by proposing a heuristic model to facilitate the study of actual and perceived cues deception, and their correspondence. This framework influenced subsequent meta-analyses and became foundational to understanding why human lie detection accuracy typically hovers around 54% correct—only slightly better than chance. The study revealed that people rely on both valid cues (reduced positivity, increased disfluencies in certain contexts) and invalid stereotypes (slow speech), leading to modest overall accuracy.

For polygraph science specifically, the findings validated the importance of verbal content analysis as a complement to physiological monitoring. The demonstration that speech content and paralinguistic features carry deception-relevant information foreshadowed modern approaches like Scientific Content Analysis (SCAN) and Criteria-Based Content Analysis (CBCA), though the modest effect sizes cautioned against over-reliance on any single verbal indicator.

Limitations & Considerations

The study's laboratory paradigm involved low-stakes deception about affective preferences rather than consequential lies about behavior or facts. This limits generalizability to forensic and screening contexts where motivations, consequences, and cognitive demands differ substantially. The 40 senders were likely college students (typical for the era), raising questions about how findings translate to diverse populations and age groups.

Additionally, the deception task required pretending to feel differently about acquaintances—a specific form of affective misrepresentation. Real-world deception often involves concealing actions, fabricating alibis, or denying knowledge, which may produce different verbal patterns. The correlational design, while sophisticated for its time, cannot establish causal mechanisms explaining why certain speech features accompany deception.

Practical Applications

For polygraph examiners, this research underscores the value of attending to verbal content while maintaining awareness of the limits of speech-based detection. The finding that perceivers judged as deceptive descriptions that were spoken slowly and contained many um's and er's, and expressions of liking that contained many other references, few self-references, and many nonspecific descriptors highlights common but sometimes misleading heuristics. Examiners should recognize that nervousness (manifesting as slow, hesitant speech) may be mistaken for deception, while more subtle content features—like shifts in language positivity or specificity—may carry diagnostic value.

The research supports integrating structured statement analysis with physiological polygraph measures rather than relying on global impressions of "nervousness" or demeanor. The demonstrated correspondence between actual and perceived cues suggests that training can enhance detection by teaching examiners to focus on empirically validated indicators while discounting misleading stereotypes. However, the modest correlations caution against overconfidence—no verbal cue or pattern provides certainty about truthfulness.