Deception Detection: Comparing the Lexicon-Based Approaches and Verifiability Approach

Deception detection methods have advanced beyond identifying nonverbal behavior with increasing focus on content analysis of verbal behavior. Content analysis deception detection has been dominated by lexicon-based approaches. The verifiability approach is a newer method of content analysis deception detection. The purpose of this study was to compare these approaches on the same data set to assess whether lexicon-based or verifiability approach content analysis is more accurate at detecting deception. In this study, lexicon-based content analysis using the Linguistic Inquiry and Word Count (LIWC) software was conducted on accounts from a prior verifiability approach study. Separate regression models were constructed to assess honest and deceptive alibi accounts using each approach. Both the lexicon-based and verifiability models detected deception at above chance accuracy; predictive accuracy did not significantly differ between these approaches. Implications for the application of content analysis to detect deception are discussed.

Secondary analysis applying LIWC lexicon-based content analysis to alibi statements from a prior verifiability approach study, with separate logistic regression models constructed for each approach to compare classification accuracy between honest and deceptive accounts.

Background & Context

Deception detection has undergone a fundamental shift over the past two decades. While traditional approaches focused heavily on nonverbal behavior—often with limited success—the field has increasingly turned to verbal content analysis as a more promising avenue for identifying deception. This evolution reflects a growing recognition that what people say, rather than how they physically present themselves, may offer more reliable indicators of truthfulness.

Content analysis deception detection has been dominated by lexicon-based approaches, but the verifiability approach represents a newer method that examines whether details in statements can be independently checked. Lexicon-based methods, particularly using the Linguistic Inquiry and Word Count (LIWC) software, analyze word categories—pronouns, emotion words, cognitive terms—to identify linguistic patterns associated with deception. The verifiability approach, by contrast, focuses on whether statements contain details that can be verified by investigators (such as witnesses, receipts, or CCTV footage).

Despite both methods enjoying empirical support, no prior study had directly compared their relative accuracy using the same dataset. This gap left practitioners without clear guidance on which approach might be more effective for real-world applications, particularly in the forensically important context of alibi evaluation.

Research Design & Methodology

This study compared lexicon-based and verifiability approaches on the same data set by applying LIWC software to alibi accounts from a prior verifiability approach study, with separate regression models constructed to assess honest and deceptive accounts. The research employed a secondary analysis design, reanalyzing verbal statements that had been collected in an earlier investigation specifically focused on the verifiability approach.

The methodology involved participants providing alibi statements about their whereabouts during a specific timeframe. Truth-tellers provided honest accounts of their actual activities, while deceivers fabricated false alibis. For the lexicon-based analysis, researchers used LIWC—a computerized text analysis program that counts words in psychologically meaningful categories including first-person pronouns, cognitive complexity words, emotion terms, and other linguistic dimensions previously linked to deception.

The statistical approach employed logistic regression modeling for each method separately, allowing researchers to determine:

• Which linguistic variables from LIWC best predicted deception versus honesty
• Which verifiability metrics (proportion of verifiable details, number of checkable facts) distinguished true from false alibis
• The overall classification accuracy of each approach
• Whether one method significantly outperformed the other in predictive validity

Results & Key Findings

Both the lexicon-based and verifiability models detected deception at above chance accuracy, but predictive accuracy did not significantly differ between these approaches. This represents a critical finding: both methods worked, but neither demonstrated clear superiority over the other.

Specific results included:

• Above-chance classification for both approaches, indicating that each method provided useful diagnostic information for distinguishing honest from deceptive alibi accounts
• No significant difference in overall predictive accuracy between lexicon-based (LIWC) and verifiability approaches
• Comparable performance across multiple statistical indicators, suggesting both methods tap into genuine markers of deception
• Evidence that different theoretical frameworks (linguistic patterns vs. checkable details) can converge on similar diagnostic outcomes

The findings support the validity of both theoretical approaches to verbal deception detection. The lexicon-based method successfully identified linguistic markers—likely including cognitive complexity, distancing language, and emotion words—that differentiated truth-tellers from liars. Simultaneously, the verifiability approach confirmed that truth-tellers provide more checkable details while liars strategically avoid providing information that can be independently verified.

Discussion & Significance

This study makes an important contribution by providing the first direct comparison of two major content analysis approaches using identical verbal statements. The finding that both methods perform comparably suggests they may be detecting deception through different but equally valid pathways. Lexicon-based approaches capture unconscious linguistic patterns reflecting cognitive load, emotional states, and self-presentation strategies. Verifiability approaches, conversely, identify strategic choices about information disclosure.

The results have important theoretical implications for understanding deception mechanisms. Rather than one approach being "correct" and the other flawed, the data suggest that deception manifests in multiple ways simultaneously—through both automatic linguistic shifts and conscious strategic information management. This aligns with dual-process models of cognition, where both controlled and automatic processes influence verbal behavior during deception.

From a practical standpoint, the comparable performance is significant. It suggests practitioners might choose between approaches based on operational constraints rather than accuracy concerns. LIWC offers speed and objectivity through automated analysis, while the verifiability approach provides a transparent rationale (checkable vs. uncheckable details) that may be more persuasive in legal contexts where algorithmic decisions face scrutiny.

Limitations & Considerations

As a secondary analysis of existing data, the study was constrained by the original research design. The alibi statements were collected under specific experimental conditions that may not fully represent the complexity of real criminal investigations, where stakes are higher and deception strategies may be more sophisticated. The sample characteristics—including the number of participants, demographic composition, and specific alibi scenarios—would influence the generalizability of findings.

Additional limitations include:

• Laboratory setting may not capture the full range of deception strategies used in high-stakes criminal contexts
• Single dataset limits generalizability across different types of deceptive scenarios beyond alibis
• Potential for both methods to be vulnerable to countermeasures if suspects become familiar with these detection techniques
• Unknown whether combining both approaches might yield superior accuracy than either alone

Practical Applications

For law enforcement and forensic practitioners, these findings suggest flexibility in choosing deception detection methods based on case-specific needs. When rapid screening of multiple statements is required, automated lexicon-based analysis offers efficiency. When building a case that must withstand legal scrutiny, the verifiability approach's focus on independently checkable facts may be more defensible in court, as it provides concrete investigative leads rather than probabilistic linguistic patterns.

The study also validates continued research investment in both approaches rather than prematurely abandoning one in favor of the other. Future applications might benefit from hybrid models that integrate linguistic analysis with verifiability coding, potentially capturing complementary aspects of deceptive communication. Practitioners should recognize that multiple tools exist for content analysis, each with distinct advantages depending on investigative context, available resources, and evidentiary standards.