The Control Question Technique: A search for improved decision

Elaad's 1999 study investigated optimal decision rules for CQT polygraph examinations using field data from Israeli Police criminal investigations. The research systematically explored how different cutoff scores and inconclusive zone boundaries affect error rates, finding that asymmetric decision rules could improve balanced accuracy by addressing the persistent problem of unequal false positive and false negative rates in operational settings.

Analysis of verified field polygraph cases from Israeli Police investigations using numerical scoring, with ground truth established through confession verification and systematic exploration of different decision rule configurations.

Elaad's 1999 study directly addressed the persistent problem of asymmetric error rates in CQT polygraph examinations by analyzing real-world field data from Israeli criminal investigations. Using the standard seven-position numerical scoring method, the research systematically explored different decision rule configurations to determine which cutoff scores and inconclusive zone widths would optimize accuracy for both innocent and guilty examinees. Results demonstrated that asymmetric decision boundaries—using different thresholds for deceptive versus non-deceptive classifications—substantially reduced error rate imbalances compared to symmetric rules. The findings revealed a critical trade-off: wider inconclusive zones reduced errors but increased indeterminate outcomes, and no single decision rule could optimize all performance metrics simultaneously. The study concluded that law enforcement agencies must explicitly prioritize their error tolerance when establishing polygraph decision rules.

The research challenges the widespread use of symmetric decision boundaries (such as ±6 scoring) in polygraph practice and demonstrates that agencies should adopt asymmetric cutoff scores tailored to their specific operational priorities and tolerance for false positives versus false negatives. This has important implications for standardizing polygraph procedures in criminal investigations, particularly for high-stakes cases where error consequences differ significantly.

Background & Context

The late 1990s marked a critical juncture in polygraph science, as practitioners and researchers grappled with optimizing decision-making rules for the Comparison Question Technique (CQT). Elaad's 1999 study emerged within a special edition of Polygraph journal dedicated entirely to chart interpretation—a signal of the field's growing emphasis on standardizing scoring methods and improving diagnostic accuracy.

A persistent challenge in CQT polygraph had been the problem of asymmetric error rates: field studies consistently showed that false positive rates (incorrectly classifying truthful examinees as deceptive) differed substantially from false negative rates (incorrectly classifying deceptive examinees as truthful). This asymmetry raised critical questions about optimal cutoff scores and the width of inconclusive zones, particularly in high-stakes criminal investigations conducted by agencies like the Israel National Police.

Elaad's research directly addressed this optimization problem using real-world field data, seeking to identify decision rules that would balance accuracy for both truthful and deceptive examinees while minimizing errors with serious consequences.

Research Design & Methodology

The study analyzed field polygraph examination data from actual criminal investigations, with outcomes published in Polygraph, Volume 28, pages 65-73. The research examined cases from Israeli Police polygraph operations, utilizing the standard seven-position numerical scoring method that was routine practice in that agency.

The methodology focused on exploring different decision rule configurations by systematically varying cutoff scores and inconclusive zone boundaries. Previous research had found asymmetric error rates in field applications, which were partially remedied by adopting asymmetrical cut scores in decision rules. Elaad's analysis tested multiple decision criteria to determine which configuration optimized accuracy across both innocent and guilty examinees.

Ground truth criteria were established through confession verification—a common but imperfect approach in field validity research. Cases were drawn from the Israel National Police Polygraph Laboratory's archive of verified examinations where subsequent confessions or other independent evidence established the actual guilt or innocence of examinees.

Results & Key Findings

The study revealed that decision rule optimization significantly impacts test accuracy, particularly in managing the trade-off between false positive and false negative errors. Asymmetric error rates were found, consistent with earlier research, and were partially remedied by adopting asymmetrical cut scores.

Key statistical findings included:

• Different cutoff thresholds produced substantially different error patterns for innocent versus guilty examinees
• Wider inconclusive zones reduced error rates but increased the proportion of indeterminate outcomes
• Asymmetric decision boundaries (different thresholds for deceptive vs. non-deceptive calls) improved balanced accuracy compared to symmetric rules
• The optimal decision rule depended on the relative costs assigned to false positive versus false negative errors

The research demonstrated that no single decision rule universally optimizes all performance metrics, requiring agencies to explicitly consider their priorities when establishing cutoff scores.

Discussion & Significance

Elaad's findings had important implications for operational polygraph practice. The demonstration that asymmetric cutoffs could reduce error rate imbalances challenged the widespread use of symmetric decision boundaries (e.g., ±6) that treated deceptive and non-deceptive calls equivalently. This was particularly relevant in criminal investigation contexts, where the consequences of false accusations differ from those of failing to detect deception.

The research also highlighted how using conservative inconclusive zones (such as ±5) might mitigate bias effects by reducing the possibility that borderline cases could shift between deception indicated and no deception indicated. This insight connected decision rule optimization to broader concerns about examiner bias and quality assurance.

The study contributed to an ongoing evolution in polygraph scoring methodology, complementing contemporaneous research on numerical scoring systems and reinforcing the importance of empirically-derived decision criteria rather than arbitrary thresholds. It provided field validation for principles that had primarily been explored in laboratory settings.

Limitations & Considerations

The reliance on confession-based ground truth introduces inherent limitations. Ground truth information established through confessions is not systematically collected and may be inadequately described, with verification sometimes by confession alone. False confessions, while rare, could contaminate the criterion measure, and innocent suspects who confess to crimes committed by co-suspects introduce verification errors.

The generalizability of optimal cutoffs derived from Israeli Police field data to other contexts (different testing formats, examiner populations, or investigative settings) remains uncertain. Base rates of guilt, case selection factors, and organizational pressures may differ substantially across jurisdictions, potentially affecting the optimal decision boundary placement.

Additionally, the study analyzed archival data rather than prospectively testing decision rules, meaning the findings reflect correlational patterns rather than experimental validation of specific scoring protocols.

Practical Applications

For polygraph examiners and agencies, this research underscores the importance of explicitly considering error costs when establishing scoring thresholds. In criminal investigations where false accusations carry severe consequences, asymmetric rules favoring false negative errors over false positives may be appropriate. Conversely, in security screening where missing a threat poses greater risk, different cutoffs would be justified.

The findings also support the use of expanded inconclusive zones as a quality control measure, particularly when cases fall near decision boundaries. Rather than viewing inconclusive results as failures, agencies should recognize them as appropriate responses to ambiguous evidence. Elaad's work demonstrates that decision rules are not merely technical details but policy choices reflecting an organization's priorities and tolerance for different error types.