A Study of the Validity of Polygraph Examinations in Criminal Investigation

Major NIJ-funded field study demonstrating CQT accuracy exceeding 90% in actual criminal investigations when conducted by skilled examiners.

NIJ-funded field study examining CQT accuracy in actual criminal investigations. Ground truth established through confessions and independent evidence. Examinations conducted by experienced examiners using standardized numerical scoring.

This major field study, funded by the National Institute of Justice, demonstrated CQT accuracy exceeding 90% when examinations were properly conducted by skilled examiners using standardized procedures. The NIJ funding added institutional credibility, and the study established a benchmark for forensic polygraph validity in actual criminal investigations that has been widely cited in subsequent reviews and legal proceedings.

NIJ-funded research carries substantial credibility in legal and policy contexts. The 90%+ accuracy rate in field conditions with actual criminal suspects provides strong support for forensic polygraph applications.

Background & Context

The 1988 Raskin study emerged at a critical juncture in polygraph history, when law enforcement increasingly relied on the technique yet scientific validation remained contested. Prior research had largely consisted of laboratory mock-crime experiments, leaving questions about whether findings could generalize to real criminal investigations. This project examined 241 confirmed and unconfirmed polygraph charts from U.S. Secret Service examinations conducted during FY 1983 and FY 1984 to determine the validity of the Control Question Technique (CQT) for assessing truth and deception in criminal investigations.

The National Institute of Justice funding (Grant No. 85-IJ-CX-0040) added substantial institutional credibility to the research and reflected federal interest in establishing evidence-based standards for forensic polygraph applications. This study represented one of the first large-scale attempts to validate CQT accuracy using actual criminal cases with independently confirmed outcomes, rather than laboratory simulations where participants faced no real consequences for detection.

The research addressed three fundamental questions that had plagued polygraph validation efforts: whether examiner characteristics affected accuracy, whether computerized scoring could match or exceed human interpretation, and critically, whether laboratory findings could predict field performance—a persistent concern for translating research into practice.

Research Design & Methodology

The project used 6 polygraph examiners and a psychophysiologist at the University of Utah to 'blindly' interpret 241 cases selected by computer program from 1,757 polygraph examinations conducted by the U.S. Secret Service at its Washington, D.C., headquarters. Ground truth was established through confessions and independent evidence that confirmed whether examinees had been truthful or deceptive, creating a verified criterion against which polygraph decisions could be measured.

The methodology incorporated multiple layers of analysis to assess different factors affecting validity. The study assessed: (1) the performance of polygraph examiners with different educational backgrounds and experience with polygraph techniques; (2) the efficacy of a computer method for interpreting the outcomes of polygraph examinations; and (3) the extent to which laboratory mock-crime experiments provide information and results that have implications for field applications.

Key design features included:

• Blind evaluation protocol where interpreters had no case information beyond the physiological charts
• Computer interpretation using algorithms developed at the University of Utah
• Standardized numerical scoring system applied across all evaluations
• Comparison of original examiner decisions with blind reinterpretations
• Analysis of laboratory versus field data patterns to assess ecological validity

Results & Key Findings

Results indicate that the accuracy of human and computer interpretations was very high, ranging from 91 percent to 96 percent correct on confirmed truthful answers, and 85 percent to 95 percent correct on confirmed deceptive answers. These findings represented some of the strongest field validation data available for the CQT at the time, demonstrating that the accuracy of the CQT can exceed 90% when properly employed by skilled examiners.

The study revealed important differences between original and blind interpretations:

• Original examiner accuracy: 91-96% correct on confirmed truthful answers and 85-95% correct on confirmed deceptive answers
• Blind interpretation accuracy: 63 percent to 85 percent on truthful answers, and 84 percent to 95 percent correct on deceptive answers
• Computer interpretation accuracy: 95 percent to 96 percent on confirmed truthful suspects, and 83 percent to 96 percent on confirmed deceptive subjects

The computer algorithms demonstrated performance comparable to or exceeding blind human interpretation, suggesting potential for objective decision support. The accuracy of human and computer interpretations was very high. Notably, original examiners who conducted the full examination process (including pretest interview and behavioral observations) achieved the highest accuracy rates, indicating that contextual information enhanced diagnostic precision.

The generalizability analysis produced particularly important findings for research methodology. Extensive analyses showed substantial similarity between data sets obtained from laboratory subjects and criminal suspects when comparable procedures are used. This suggested that well-designed laboratory experiments could inform field practice, though adjustments in decision criteria should be made to reduce false positive errors caused by overprediction of deceptive field outcomes using criteria derived from laboratory experiments.

Discussion & Significance

The study's convergence of high accuracy rates across multiple evaluation methods provided robust support for CQT validity in field applications. The finding that computerized analysis rivaled or exceeded blind human scoring suggested algorithmic approaches could enhance objectivity and consistency in polygraph interpretation—a significant advancement for standardizing forensic practice. The original examiners' superior performance highlighted the value of comprehensive examination protocols beyond chart analysis alone.

The laboratory-to-field generalizability findings had profound implications for polygraph research methodology. By demonstrating that laboratory and field data showed substantial similarity under comparable conditions, the research validated the use of controlled experiments to study underlying psychophysiological processes while cautioning that decision thresholds required calibration for different contexts. This bridged a longstanding divide between critics who dismissed laboratory research as artificial and practitioners who relied primarily on field experience.

The NIJ funding and Secret Service partnership lent the research exceptional credibility in legal and policy contexts. The study became widely cited in court proceedings, government reviews, and professional training as evidence for polygraph validity under proper conditions. It established benchmarks against which subsequent field studies would be compared and influenced the development of quality control standards emphasizing examiner training, standardized protocols, and computer-assisted decision support.

Limitations & Considerations

The study's reliance on Secret Service cases created potential selection bias, as these examinations involved federal investigations with potentially different characteristics than local law enforcement cases. The confirmed case sample represented a subset of all examinations conducted, and ground truth verification through confessions may have excluded cases where innocent examinees were falsely accused but unable to prove their innocence through alternative evidence.

The blind interpretation methodology, while scientifically rigorous, removed contextual information available to original examiners—information that may be legitimately diagnostic in practice. The lower accuracy of blind interpretations raises questions about whether chart features alone provide sufficient information or whether integrating behavioral observations and case facts enhances rather than contaminates clinical judgment. The study did not examine specific error patterns or identify case characteristics associated with false positives versus false negatives.

The findings indicated that adjustments in decision criteria should be made to reduce false positive errors when translating laboratory-derived standards to field settings, but the research did not specify optimal threshold values. This leaves examiners with general guidance rather than precise decision rules for balancing sensitivity and specificity in real-world applications.

Practical Applications

This research supports several evidence-based practices for forensic polygraph examination. The 90%+ accuracy rates achieved by skilled examiners using standardized procedures demonstrate that CQT can provide valuable investigative information when properly conducted—though not at levels justifying sole reliance for criminal adjudication. The computer algorithm performance suggests that automated scoring systems can serve as quality control tools, flagging charts where human and computer interpretations diverge for additional review.

For consumers and defendants, the findings underscore that examiner qualifications and protocol standardization significantly impact accuracy. Polygraph results should be interpreted with awareness that even high accuracy rates mean some proportion of errors, with different error types (false positives versus false negatives) carrying different consequences. The research supports viewing polygraph as an investigative aid within a broader evidence framework rather than a definitive truth-determination method, consistent with its typical exclusion from most courtroom proceedings.