Algorithm

What Is a Polygraph Scoring Algorithm?

In polygraph science, an algorithm is a defined set of mathematical rules and statistical models applied to the physiological data recorded during a lie detector test to produce a numerical score, probability estimate, or classification decision. Modern polygraph algorithms analyse patterns across respiratory, electrodermal, and cardiovascular channels to determine whether the examinee's physiological responses are more consistent with deception or truthfulness.

Algorithms represent one of the most significant advances in polygraph science over the past three decades, transforming the field from a largely subjective practice into one grounded in quantitative statistical analysis.

How Polygraph Scoring Algorithms Work

Polygraph algorithms follow a systematic multi-step process:

• Feature extraction — The algorithm identifies and measures specific response characteristics (called Kircher features) from the raw physiological data. These include EDA amplitude, respiratory line length (RLL), cardiovascular baseline changes, and pulse amplitude variations
• Comparison scoring — Extracted features from responses to relevant questions are mathematically compared against features from responses to comparison questions within each chart
• Statistical classification — The combined feature scores are input into a statistical model (logistic regression, discriminant analysis, or Bayesian analysis) that calculates a probability of deception
• Decision application — The calculated probability or score is compared against empirically derived cut-scores to classify the result as DI (Deception Indicated), NDI (No Deception Indicated), or INC (Inconclusive)

Major Validated Polygraph Algorithms

Several validated polygraph algorithms are in widespread use today:

OSS-3 (Objective Scoring System, Version 3)

Developed by Nelson, Krapohl, and Handler, OSS-3 is a free, open-source algorithm using logistic regression trained on confirmed field polygraph cases. It is one of the most extensively validated algorithms available, with research demonstrating accuracy rates exceeding 90%. OSS-3's open-source nature makes it accessible to any examiner or researcher and allows independent verification of its methodology.

PolyScore

A proprietary algorithm developed at Johns Hopkins University Applied Physics Laboratory, PolyScore uses linear discriminant analysis and Bayesian probability to classify examinees. It is integrated into Lafayette Instrument Company polygraph platforms and is widely used by U.S. federal agencies.

ESS-M (Empirical Scoring System – Modified)

An automated implementation of the manual Empirical Scoring System (ESS), built into Limestone Technologies instruments. Uses empirically derived regression weightings for multi-channel data analysis.

CPS Elite

A computerised scoring system developed by Scientific Assessment Technologies based on University of Utah research, integrated into their polygraph software platform.

Why Algorithms Are Important for Polygraph Accuracy

Automated scoring algorithms address one of the key historical challenges in polygraph testing: subjectivity and interrater variability. When human examiners manually score polygraph charts, results can vary between different scorers — a problem measured by interrater reliability. Algorithms apply identical mathematical rules to every examination, producing consistent results regardless of who runs the analysis.

Research consistently shows that validated algorithms perform at accuracy levels comparable to or exceeding the average of experienced human scorers, while also achieving significantly higher interrater reliability (effectively perfect, since the same algorithm always produces the same result for the same data).

Limitations of Polygraph Algorithms

Algorithms are only as good as the data they receive. Several factors can degrade algorithm performance:

• Poor data quality — Improperly attached sensors, excessive artifacts, or non-standard testing conditions
• Countermeasure attempts — Deliberate manipulation by the examinee, though activity sensors help detect these
• Overfitting risk — Algorithms must be properly cross-validated on independent datasets to ensure they generalise beyond training data
• Technique mismatch — An algorithm validated on one testing technique may not perform optimally when applied to a different technique format

For the latest research on polygraph scoring algorithms and accuracy, visit our research database or explore the Polygraph Examiner Hub.