Comparison of OSS-3 and PolyScore in Multi-Issue and Single-Issue Field Exams

Comparison of OSS-3 and PolyScore computerized scoring algorithms on multi-issue and single-issue field examinations.

Comparative analysis of two scoring algorithms applied to field examination data with confirmed outcomes.

Krapohl and Dutton compared OSS-3 and PolyScore algorithms, finding OSS-3 correctly identified 71% truthful and 96% deceptive with a 14% inconclusive rate, outperforming PolyScore on multiple-issue tests. The comparison helps practitioners select the most appropriate scoring algorithm for their testing context.

Algorithm comparison data supports evidence-based selection of scoring methods and quality assurance monitoring.

Background & Context

The evolution of computerized scoring algorithms represents one of the most significant advances in modern polygraph practice. For decades, the field relied on subjective human interpretation of physiological charts, which introduced variability and potential bias into deception detection. The 2024 comparison by Krapohl and Dutton addresses a critical gap in practitioner knowledge: which automated scoring algorithm performs better with different types of field examinations.

This research builds on a foundation established by earlier algorithm comparisons, including the seminal work by Dollins, Krapohl, and Dutton in 2000 that evaluated multiple computerized scoring systems. OSS-3 was developed as an independent collaborative project to provide a robust, widely applicable, and mathematically sound computerized polygraph scoring model, while PolyScore emerged from different statistical foundations. The question of which algorithm performs optimally across different testing contexts—particularly multi-issue versus single-issue examinations—has direct implications for examiner decision-making and quality assurance protocols.

Research Design & Methodology

The study examined 48 single-issue examinations and multiple-issue field examinations, applying both the OSS-3 and PolyScore algorithms to the same confirmed dataset. This comparative design allowed for direct assessment of each algorithm's performance characteristics under identical conditions. The research utilized field examination data with verified outcomes, providing ecological validity that laboratory studies cannot offer.

The methodology involved scoring confirmed cases where ground truth was established through confessions, corroborating evidence, or other independent verification methods. Both algorithms processed the same physiological data, eliminating potential confounds from different data collection procedures. Key methodological elements included:

• Confirmed field cases with known outcomes (truthful vs. deceptive)
• Parallel scoring by both OSS-3 and PolyScore algorithms
• Separate analysis for multi-issue and single-issue test formats
• Classification of results into truthful, deceptive, and inconclusive categories

Results & Key Findings

OSS-3 substantially outperformed PolyScore on multi-issue examinations. The OSS-3 algorithm correctly identified 71% of truthful examinees and 96% of deceptive examinees, with a 14% inconclusive rate. This balanced performance demonstrates strong sensitivity to deception while maintaining reasonable specificity for truthful individuals.

In stark contrast, PolyScore's multi-issue algorithm correctly identified only 45% of truthful subjects and 87% of deceptive subjects, but had a 43% inconclusive rate. The dramatically elevated inconclusive rate for PolyScore represents a critical limitation, as inconclusive results provide no actionable information and often necessitate retesting.

Performance metrics for multi-issue field examinations:

• OSS-3 truthful detection: 71% vs. PolyScore: 45%
• OSS-3 deceptive detection: 96% vs. PolyScore: 87%
• OSS-3 inconclusive rate: 14% vs. PolyScore: 43%
• OSS-3 demonstrated superior balanced accuracy across both truthful and deceptive classifications

Discussion & Significance

These findings have profound implications for algorithm selection in operational polygraph settings. The superior performance of OSS-3 correlates favorably with other research validating its effectiveness. The algorithm's ability to correctly classify 96% of deceptive examinees while maintaining a low inconclusive rate makes it particularly valuable for investigative and screening contexts where identifying deception is paramount.

The performance differential is especially striking for multi-issue examinations, which are commonly used in pre-employment screening and periodic law enforcement testing. Multiple-issue testing has long been recognized as more challenging than single-issue examinations, with some experts noting that shortcuts erode validity. OSS-3's superior performance on these complex test formats suggests it may be better equipped to handle the statistical challenges inherent in analyzing responses across multiple relevant questions.

The dramatic difference in inconclusive rates between the two algorithms—14% versus 43%—represents a critical practical consideration. Computer scoring algorithms provide objective and reliable evaluation of physiological data, independent of examiner knowledge or expectations. However, an algorithm that produces inconclusive results in nearly half of cases provides limited operational value, increasing costs and examiner workload through required retesting.

Limitations & Considerations

While the study provides valuable comparative data, several limitations warrant consideration. The research utilized field examination data, which inherently contains more variability than laboratory studies but better represents real-world conditions. Computers cannot distinguish high-quality data from low-quality atypical data or detect countermeasures, whereas human examiners are better at evaluating data quality. This suggests that optimal polygraph practice combines algorithmic scoring with examiner oversight.

The study focused specifically on multi-issue and single-issue examinations but did not examine all possible test formats or question structures. Additionally, algorithm performance can be influenced by factors including examiner technique, equipment calibration, and examinee physiological responsivity. The confirmed outcomes methodology, while robust, may not capture the full spectrum of cases encountered in operational settings where ground truth is often ambiguous.

Practical Applications

This research provides evidence-based guidance for practitioners selecting computerized scoring algorithms. For agencies conducting multi-issue screening examinations, OSS-3 appears to offer substantial advantages in both detection accuracy and operational efficiency. OSS-3 is proposed as a validated option that provides perfect reliability through computerized analysis, making it particularly suitable for quality assurance and standardization efforts.

The findings support a hybrid approach combining algorithmic objectivity with human expertise. Examiners should pair their evaluation of data quality—identifying artifacts and countermeasures—with computer algorithms that provide unbiased numerical analysis. This integration leverages the strengths of both approaches: human judgment for data quality and contextual factors, and computerized scoring for objective, reproducible analysis. Agencies should consider algorithm performance characteristics when establishing testing protocols and quality control procedures.