Diagnosing deceptive and mixed messages from verbal and nonverbal cues

Background & Context

This 1982 study by DePaulo, Rosenthal, Green, and Rosenkrantz represents a foundational contribution to understanding how verbal and nonverbal channels convey deceptive and emotionally complex messages. The research emerged during a period of intense scientific interest in nonverbal communication and deception detection, building on earlier work by Ekman and Friesen on "nonverbal leakage" and the theoretical proposition that different communication channels might reveal truth differently.

The study addressed a critical gap in deception research: most prior work had focused exclusively on detecting outright lies, but real-world communication often involves mixed emotional states—ambivalence and indifference—that might produce similar cues to deception. Understanding whether judges could distinguish truly deceptive messages from emotionally complex honest messages was essential for both theoretical development and practical application.

The research also tested competing theories about the relative "leakiness" of verbal versus nonverbal channels, with particular relevance to Paul Ekman's hypothesis that the body and voice might leak true feelings more readily than carefully controlled facial expressions and words.

Research Design & Methodology

Forty senders created three types of videotaped messages: "pure" messages describing people they genuinely liked or disliked; "mixed" messages about people they felt ambivalent about or indifferent toward; and "deceptive" messages where they described liked persons while pretending to dislike them, and vice versa. This design allowed comparison of outright deception against both straightforward and emotionally complex honest communication.

Eighty judge-subjects evaluated these messages under five experimental conditions: tone of voice only (content-filtered speech), visual cues only, verbal transcripts only, undistorted audio, and full audiovisual presentation. The content-filtering technique removed semantic information while preserving vocal qualities, allowing assessment of purely nonverbal vocal cues. The last three conditions contained complete verbal information, while the first two were entirely nonverbal.

Key methodological features included:

• Counterbalanced presentation orders to control sequence effects
• Multiple rating dimensions including deceptiveness, ambivalence, discrepancy, indifference, and tension
• Systematic comparison of verbal-rich versus nonverbal-only channels
• Use of both detection accuracy and dimensional judgment measures

Results & Key Findings

Judges perceived deceptive messages as significantly more ambivalent, discrepant, deceptive, indifferent, and tense than pure messages, with discriminations along the ambivalence dimension actually stronger than discriminations based on deceptiveness ratings. This revealed that deception produced a distinctive "mixed message" quality that observers could detect.

However, subjects did not perceive the deceptive messages as any more deceptive than the genuinely mixed (ambivalent/indifferent) messages—a critical finding showing honest ambivalence mimics deception cues.

The study revealed striking modality effects:

• Subjects with access to verbal information made more pronounced discriminations across more dimensions than those with only nonverbal cues, and verbally based discriminations were usually more accurate
• When perceivers judged speakers' underlying affects while speakers were lying, their verbally based discriminations were very inaccurate—suggesting verbal channels effectively masked true feelings during deception
• Nonverbal channels showed some diagnostic value but were generally less informative than verbal content

The data provided only partial support for each of three major theories of verbal and nonverbal encoding, suggesting deception involves more complex channel dynamics than simple models predicted.

Discussion & Significance

This research made several enduring contributions to deception science. It demonstrated that the challenge of lie detection partly stems from the similarity between deception cues and cues from genuine emotional complexity. The finding that ambivalent messages looked deceptive has important implications: not every "suspicious" signal indicates lying—it may reflect authentic inner conflict.

The verbal channel findings contradicted simplistic assumptions that nonverbal cues always "leak" truth more reliably. Instead, verbal content proved highly diagnostic for some judgments but particularly misleading when trying to infer underlying emotions during deception. This nuanced pattern suggested skilled deceivers can control their words effectively even while experiencing detectable tension.

The study became highly influential, cited extensively in Bella DePaulo's later landmark 2003 meta-analysis of deception cues and in theoretical work on multi-channel communication. It established methodological standards for deception research, including the use of content filtering and systematic comparison across communication modalities. The research also highlighted that deception detection accuracy depends critically on what judges are trying to assess—detecting that something is "off" differs from identifying the specific nature of that discrepancy.

Limitations & Considerations

The laboratory paradigm involved relatively low-stakes deception about interpersonal feelings, which may not generalize to high-consequence lies about actions, crimes, or security threats. The senders knew they were being videotaped and evaluated, potentially affecting naturalness. The study examined only one type of deception (reversed affect about persons), leaving questions about detection of other lie types.

The finding that judges' discriminations along the ambivalence dimension were stronger than discriminations based on deceptiveness, and they did not perceive deceptive messages as more deceptive than mixed messages, raises questions about whether experimental instructions adequately oriented judges to the detection task. The reliance on college students as both senders and judges limits generalizability to populations with different communication styles or deception experience.

Practical Applications

For deception detection practitioners, this research offers a crucial cautionary message: behavioral signs of tension, discrepancy, or ambivalence should not be automatically interpreted as deception indicators. Honest individuals experiencing conflicted emotions may display remarkably similar cues. This insight supports modern polygraph practice emphasizing question construction that minimizes emotional complexity in truthful responses.

The finding that verbal content provides diagnostic information—but can mislead when assessing underlying emotions—suggests multi-channel assessment strategies should weight different modalities depending on the judgment goal. When evaluating factual claims, verbal content deserves attention; when assessing emotional genuineness, nonverbal cues may prove more informative. The research underscores why validated psychophysiological measures, which bypass conscious verbal control, remain valuable complements to behavioral observation in credibility assessment.