Before the first question, base rates already shape the odds. This Bayesian guide explains how prior probabilities inform the interpretation of a lie detector test.
Prior probabilities form the mathematical backbone of modern polygraph interpretation. By integrating base rate data with test results through Bayesian methods, examiners move beyond subjective judgment to produce quantified, defensible assessments of classification confidence — dramatically improving the reliability of polygraph conclusions.
TL;DR — The Short Version
- Prior probabilities are statistical estimates of deception likelihood established before a polygraph test, based on population base rates and contextual information.
- Bayesian inference combines prior probabilities with polygraph test data to produce refined posterior probabilities, improving classification accuracy.
- Base rates vary dramatically by context — screening populations differ fundamentally from criminal investigation contexts, and using the wrong base rate compromises validity.
- Likelihood ratios quantify how strongly polygraph results support one hypothesis over another, serving as the multiplier that updates prior probabilities.
- Properly calibrated priors reduce both false positive and false negative rates, producing more defensible polygraph conclusions.
- The APA meta-analysis found 89% accuracy for event-specific diagnostic tests and 85% for multi-issue screening — these figures provide the foundation for Bayesian calculations.
Who This Guide Is For
- Polygraph examiners seeking to improve the statistical rigor of their interpretations and reduce classification errors
- Polygraph training students learning advanced test interpretation and scoring methodology
- Quality assurance professionals evaluating polygraph programs and examiner performance
- Attorneys and legal professionals who need to understand the statistical basis of polygraph conclusions
- Researchers in psychophysiology and deception detection studying Bayesian applications
- Law enforcement supervisors overseeing polygraph units and evaluating testing protocols
What Are Prior Probabilities in Polygraph Testing?
Definition and Origins
In polygraph testing, prior probabilities represent the best statistical estimate of the likelihood that a given subject is deceptive or truthful before any polygraph data is collected. These estimates are not arbitrary guesses — they are derived from empirical data about the population being tested, the nature of the examination, and the context surrounding the referral.
The concept originates from Bayesian probability theory, a branch of mathematics developed by Thomas Bayes, an English statistician, philosopher, and Presbyterian minister born around 1701 [1]Verified Thomas Bayes — Wikipedia
Confirms Thomas Bayes was born c. 1701, was an English statistician, philosopher, and Presbyterian minister. His work was published posthumously in 1763 by Richard Price.. Bayes never published what would become his most famous accomplishment; his notes were edited and published posthumously by Richard Price, who presented the work to the Royal Society in 1763 [1]Verified Thomas Bayes — Wikipedia
Confirms Thomas Bayes was born c. 1701, was an English statistician, philosopher, and Presbyterian minister. His work was published posthumously in 1763 by Richard Price.. The resulting essay, 'An Essay Towards Solving a Problem in the Doctrine of Chances,' established the mathematical foundation for updating beliefs based on new evidence [1]Verified Thomas Bayes — Wikipedia
Confirms Thomas Bayes was born c. 1701, was an English statistician, philosopher, and Presbyterian minister. His work was published posthumously in 1763 by Richard Price..
In Bayesian reasoning, every analysis begins with a 'prior' — an initial probability estimate that is subsequently updated as new evidence becomes available. In polygraph testing, the 'new evidence' is the physiological data collected during the examination. Understanding how examiners evaluate polygraph data is essential background for appreciating how this physiological evidence feeds into Bayesian calculations.
Raymond Nelson and Bayesian Analysis in Polygraph Practice
Raymond Nelson, a past President of the American Polygraph Association and a prominent researcher in the field [2]Verified Raymond Nelson Research Profile
Confirms Raymond Nelson is a polygraph researcher who has published extensively on Bayesian analysis, prior probabilities, posterior probabilities, ESS-M scoring, and APA standards including Standard 1.8.3., has been instrumental in advancing the application of Bayesian methods to polygraph testing. In the APA Magazine's January/February 2016 issue (Volume 49, Issue 1), Nelson explored how prior probabilities and base rates interact with test accuracy to affect confidence in polygraph outcomes [3]Verified Base Rates and Prior Probabilities in Polygraph Testing — APA Magazine 2016, Vol. 49, Issue 1
Confirms Nelson's 2016 article on base rates, prior probabilities, and their interaction with test accuracy in the APA Magazine January/February 2016 issue.. His analysis demonstrated that even with a highly accurate test, low or high base rates can markedly affect the confidence in the test result [3]Verified Base Rates and Prior Probabilities in Polygraph Testing — APA Magazine 2016, Vol. 49, Issue 1
Confirms Nelson's 2016 article on base rates, prior probabilities, and their interaction with test accuracy in the APA Magazine January/February 2016 issue..
Nelson and Finley Turner subsequently published a detailed paper on 'Bayesian probabilities of deception and truth-telling for single and repeated polygraph examinations,' which demonstrated that for any probabilistic test result, the posterior likelihood of deception or truth-telling varies in mathematically predictable ways in response to the prior probability [4]Verified Bayesian Probabilities of Deception and Truth-Telling for Single and Repeated Polygraph Examinations
Confirms Nelson and Turner's research on Bayesian posterior probabilities for single and repeated polygraph exams, including results for various prior probability distributions.. This work showed that Bayesian analysis is both intuitive and practical for interpreting polygraph results, including applications to repeated testing strategies [4]Verified Bayesian Probabilities of Deception and Truth-Telling for Single and Repeated Polygraph Examinations
Confirms Nelson and Turner's research on Bayesian posterior probabilities for single and repeated polygraph exams, including results for various prior probability distributions..
Nelson also contributed the scientific (analytic) theory of polygraph testing, published in APA Magazine Volume 49, Issue 5, which established the theoretical basis for understanding why physiological reactions differ between deceptive and truthful responses to test questions [5]Verified Scientific (Analytic) Theory of Polygraph Testing — APA Magazine 2016, Vol. 49, Issue 5
Confirms Nelson (2016) published the scientific analytic theory of polygraph testing in APA Magazine 49(5), 69-82.. Together, these contributions demonstrate why Bayesian methods have become increasingly central to modern polygraph interpretation. For a deeper comparison of statistical methods in polygraph work, see our guide on Bayesian vs. Frequentist statistics in polygraph testing.
Why Prior Probabilities Matter: The Base Rate Fallacy
Understanding the Base Rate Problem
One of the most common errors in probabilistic reasoning — and one that directly affects polygraph interpretation — is the base rate fallacy. This occurs when decision-makers ignore the prior probability of an event and focus exclusively on the diagnostic accuracy of a test. Nelson (2016) specifically emphasized this problem in the context of polygraph screening, noting that a review of the prior probability of guilt can help examiners think through the test result with the referring investigator or attorney [3]Verified Base Rates and Prior Probabilities in Polygraph Testing — APA Magazine 2016, Vol. 49, Issue 1
Confirms Nelson's 2016 article on base rates, prior probabilities, and their interaction with test accuracy in the APA Magazine January/February 2016 issue..
Consider a hypothetical polygraph test with 90% sensitivity (correctly identifies 90% of deceptive subjects) and 90% specificity (correctly identifies 90% of truthful subjects). The practical meaning of a 'deceptive' result depends entirely on the base rate of deception in the tested population [3]Verified Base Rates and Prior Probabilities in Polygraph Testing — APA Magazine 2016, Vol. 49, Issue 1
Confirms Nelson's 2016 article on base rates, prior probabilities, and their interaction with test accuracy in the APA Magazine January/February 2016 issue..
High base rate scenario (50% deceptive): If 1,000 people are tested and 500 are genuinely deceptive, the test correctly identifies 450 deceptive individuals (true positives) but falsely flags 50 truthful individuals (false positives). Among all 500 positive results, 450/500 = 90% are actually deceptive. The positive predictive value is 90%.
Low base rate scenario (5% deceptive): If 1,000 people are tested and only 50 are genuinely deceptive, the test correctly identifies 45 but falsely flags 95 truthful individuals. Among all 140 positive results, only 45/140 = 32% are actually deceptive. The positive predictive value drops to 32%.
This dramatic difference — 90% versus 32% positive predictive value — occurs despite the test having identical accuracy in both scenarios. The only variable that changed was the prior probability of deception.
Implications for Polygraph Practice
The National Research Council's 2003 report, 'The Polygraph and Lie Detection,' highlighted the base rate problem as a critical concern for screening applications [6]Verified The Polygraph and Lie Detection — National Research Council
Confirms the NRC 2003 report's analysis of base rate effects on polygraph screening, including the finding that low base rates produce high false positive rates even with accurate tests.. The NRC noted that when the base rate of deception is very low — such as one in 1,000 for security screening — even a test that correctly identifies 80% of deceptive examinees would incorrectly classify at least 100 nondeceptive individuals as deceptive for each security threat correctly identified [6]Verified The Polygraph and Lie Detection — National Research Council
Confirms the NRC 2003 report's analysis of base rate effects on polygraph screening, including the finding that low base rates produce high false positive rates even with accurate tests.. This underscores why examiners must understand base rates to properly communicate the meaning of their findings.
The APA's 2011 meta-analytic survey, encompassing 38 studies, 3,723 examinations, and 11,737 scored results, reported event-specific diagnostic accuracy of 89% (confidence interval 83-95%) and multi-issue screening accuracy of 85% (confidence interval 77-93%) [7]Verified Meta-Analytic Survey of Criterion Accuracy of Validated Polygraph Techniques
Confirms 38 studies, 3,723 examinations, 11,737 scored results. Event-specific accuracy: 89% (CI: 83-95%). Multi-issue screening: 85% (CI: 77-93%). Overall: 87% (CI: 80-94%).. These accuracy figures provide the foundation for Bayesian calculations, but their practical predictive value depends heavily on the prior probability applicable to each testing context.
Examiners who want to understand how polygraph results are scored and interpreted should also review our guide on evaluating polygraph data, as the numerical scoring system directly feeds into the Bayesian framework described here. Factors that can contaminate results — explored in our article on 5 things that contaminate polygraph exam results — also affect the reliability of Bayesian calculations.
Bayes' Theorem and Its Application to Polygraph Testing
The Mathematical Formula
Bayes' theorem provides the mathematical formula for updating prior probabilities with new evidence to produce posterior probabilities. The theorem is expressed as:
P(D|E) = [P(E|D) x P(D)] / P(E)
Where: - P(D|E) = Posterior probability of deception given the evidence (what we want to know) - P(E|D) = Probability of observing the test evidence if the subject is deceptive (sensitivity/true positive rate) - P(D) = Prior probability of deception (base rate) - P(E) = Total probability of observing the evidence (under all hypotheses combined)
For polygraph examiners, this formula translates into a practical decision-making tool. The prior probability P(D) represents what we know about the likelihood of deception before the test. The test data provides the evidence term P(E|D). The resulting posterior probability P(D|E) tells us how confident we can be in a deception conclusion after integrating both sources of information.
Steven Rigdon's 2018 study, 'Exact Bayesian Inference for Assessing the Accuracy of Polygraph Testing,' published in the Journal of the Indian Society for Probability and Statistics, provided a rigorous Bayesian framework specifically designed for evaluating polygraph test accuracy [8]Verified Exact Bayesian Inference for Assessing the Accuracy of Polygraph Testing
Confirms Rigdon (2018) developed a formal Bayesian framework for assessing polygraph accuracy using posterior distributions.. This research demonstrated how Bayesian methods can be formally applied to obtain posterior distributions for key polygraph performance metrics.
The Odds Form: A Practical Alternative
An alternative and often more intuitive way to apply Bayes' theorem in polygraph practice uses the odds form:
Posterior Odds = Prior Odds x Likelihood Ratio
This means the odds of deception after the test equal the odds before the test, multiplied by the likelihood ratio derived from the test results. The odds form is particularly useful because it allows examiners to work with likelihood ratios — a single number that captures how much the test evidence should shift beliefs about deception.
A likelihood ratio of 1 means the evidence is equally likely under both hypotheses and provides no diagnostic value. A ratio greater than 1 supports the deception hypothesis, while a ratio less than 1 supports truthfulness. In practical terms, if an examiner begins with even odds (50/50, or 1:1) and the polygraph test produces a likelihood ratio of 15 in favour of deception, the posterior odds become 15:1 — meaning deception is 15 times more likely than truthfulness. Converting back to probability: 15/16 = approximately 94% posterior probability of deception.
This calculation framework is also relevant to understanding PolyScore and CPS automated scoring systems, which incorporate probabilistic calculations into their decision algorithms.
Understanding Base Rates Across Polygraph Testing Contexts
Pre-Employment Screening
Base rates of deception in pre-employment screening contexts are typically low because most applicants are truthful about the issues being tested. Large applicant pools dilute the proportion of deceptive individuals. The OTA (Office of Technology Assessment) noted that a typical screening situation might involve very low base rates — in some security screening scenarios, as few as 1 guilty person per 1,000 screened [9]Verified Scientific Validity of Polygraph Testing: OTA Report — Chapter 7
Confirms the OTA's analysis of base rate effects in screening contexts, including the example of 1 guilty person per 1,000 screened and resulting false positive rates.. For law enforcement applicant screening, research using Information Gain analysis peaked in information value at a base rate of deception of approximately 19% [10]Verified Pre-Employment Polygraph Screening
Confirms base rate problem in pre-employment screening, inconclusive result handling, and Information Gain analysis showing peak information value at base rate of 19%.. This relatively low prior means false positives are a significant concern in screening contexts.
Nelson (2016) emphasized that when an agency asks screening questions that probably have a low base rate, a passed test is substantially more informative than a failed test [3]Verified Base Rates and Prior Probabilities in Polygraph Testing — APA Magazine 2016, Vol. 49, Issue 1
Confirms Nelson's 2016 article on base rates, prior probabilities, and their interaction with test accuracy in the APA Magazine January/February 2016 issue.. Understanding this asymmetry is critical for examiners working in pre-employment contexts and for agencies making hiring decisions based on polygraph outcomes.
Criminal-Specific Investigations
Criminal-specific investigations typically have higher base rates of deception because subjects are tested due to specific reasons to believe deception may be present. Nelson (2016) noted that criminal investigators usually refer suspects to polygraph who they feel are more likely than not to be guilty of a crime [3]Verified Base Rates and Prior Probabilities in Polygraph Testing — APA Magazine 2016, Vol. 49, Issue 1
Confirms Nelson's 2016 article on base rates, prior probabilities, and their interaction with test accuracy in the APA Magazine January/February 2016 issue.. In his worked examples, he demonstrated the effects of a 90% base rate of guilt in investigative contexts [3]Verified Base Rates and Prior Probabilities in Polygraph Testing — APA Magazine 2016, Vol. 49, Issue 1
Confirms Nelson's 2016 article on base rates, prior probabilities, and their interaction with test accuracy in the APA Magazine January/February 2016 issue..
The prior in these contexts is often closer to 50% or higher, providing a more balanced starting point and higher diagnostic efficiency for the test. The APA meta-analysis found that event-specific (single issue) diagnostic tests produced an aggregated accuracy of 89% (CI: 83-95%) [7]Verified Meta-Analytic Survey of Criterion Accuracy of Validated Polygraph Techniques
Confirms 38 studies, 3,723 examinations, 11,737 scored results. Event-specific accuracy: 89% (CI: 83-95%). Multi-issue screening: 85% (CI: 77-93%). Overall: 87% (CI: 80-94%)., which represents the accuracy figures most directly applicable to criminal investigation Bayesian calculations. Multiple independent studies have consistently reported polygraph accuracy exceeding 90% in well-conducted criminal investigations [11]Verified Review of Polygraph Accuracy Research
Confirms Honts and Peterson reported polygraph accuracy exceeding.90, consistent with other major accuracy reviews.[12]Verified Accuracy of Polygraph Techniques
Confirms Raskin and Podlesny reported polygraph accuracy exceeding.90 using the CQT format..
Post-Conviction Sex Offender Testing (PCSOT)
Post-conviction sex offender testing presents unique base rate challenges. The APA Model Policy defines several PCSOT examination types, including instant-offense exams, sex history exams, and maintenance exams [2]Verified Raymond Nelson Research Profile
Confirms Raymond Nelson is a polygraph researcher who has published extensively on Bayesian analysis, prior probabilities, posterior probabilities, ESS-M scoring, and APA standards including Standard 1.8.3.. Monitoring examinations may have varying base rates depending on supervision intensity, while disclosure examinations typically have higher base rates as they explore undisclosed history.
Nelson specifically noted that in PCSOT or public safety screening contexts, the base rate represents the proportion of examinees who are lying to one or more test questions, which is still the prior probability of guilt [3]Verified Base Rates and Prior Probabilities in Polygraph Testing — APA Magazine 2016, Vol. 49, Issue 1
Confirms Nelson's 2016 article on base rates, prior probabilities, and their interaction with test accuracy in the APA Magazine January/February 2016 issue.. Examiners working in PCSOT must be particularly attentive to base rate considerations, as the testing context significantly affects the expected prevalence of deception and the appropriate prior probability.
Private and Domestic Testing
Private tests for infidelity or family matters have highly variable base rates that depend on the specific circumstances of the referral. In cases with strong circumstantial evidence, the prior may be elevated; in exploratory cases, it may be closer to chance. Understanding these contextual differences is critical — an examiner conducting pre-employment screening faces a fundamentally different statistical landscape than one conducting a private lie detector test related to a specific allegation.
Likelihood Ratios: The Bridge Between Prior and Posterior
How Likelihood Ratios Work
While prior probabilities set the starting point for Bayesian analysis, likelihood ratios are the mechanism by which polygraph test data updates those priors. A likelihood ratio (LR) compares the probability of observing a particular test result under two competing hypotheses:
Likelihood Ratio = P(Test Result | Deceptive) / P(Test Result | Truthful)
A LR of 10 means the observed test data is 10 times more likely if the subject is deceptive than if truthful. A LR of 0.1 means the data is 10 times more likely if the subject is truthful. A LR of 1 means the data is equally likely under both hypotheses — the test provides no diagnostic information.
In polygraph practice, likelihood ratios are derived from the test's sensitivity and specificity — or more precisely, from the distribution of test scores among known deceptive and known truthful individuals in validated research [7]Verified Meta-Analytic Survey of Criterion Accuracy of Validated Polygraph Techniques
Confirms 38 studies, 3,723 examinations, 11,737 scored results. Event-specific accuracy: 89% (CI: 83-95%). Multi-issue screening: 85% (CI: 77-93%). Overall: 87% (CI: 80-94%).. Different numerical scores correspond to different likelihood ratios, with strongly positive or negative scores having the highest diagnostic power.
Likelihood Ratios and Score Magnitude
The concept of likelihood ratios helps explain why some polygraph scores carry more diagnostic weight than others. A strongly positive score on a validated scoring system will have a very high likelihood ratio in favour of truthfulness, while a score near the decision threshold will have a likelihood ratio closer to 1.
This framework also has direct implications for interpreting inconclusive results. An inconclusive outcome — where the numerical score falls within an ambiguous range — corresponds to a likelihood ratio near 1, meaning the test data provides little information for updating the prior. In such cases, the posterior probability remains close to the prior, and the examiner appropriately acknowledges that the test has not resolved the question. Nelson noted that in pre-employment screening, an inconclusive result cannot deliver a result that is much more predictive than simply using the base rate as a guide, making re-testing the best practice [10]Verified Pre-Employment Polygraph Screening
Confirms base rate problem in pre-employment screening, inconclusive result handling, and Information Gain analysis showing peak information value at base rate of 19%..
Honts and Schweinle (2009) adapted Information Gain analysis procedures for use with the Comparison Question Test (CQT) and its three levels of outcome — deceptive, truthful, and inconclusive — providing a Bayesian-based approach to describe the impact of base rates on information provided by polygraph procedures [13]Verified A Comprehensive Meta-Analysis of the Comparison Question Polygraph Test
Confirms Honts and Schweinle's Information Gain analysis adapted for CQT and the Bayesian-based approach to describing base rate impact on polygraph outcomes.. For more on the CQT framework, see our guide to the You-Phase Zone Comparison Test.
From Prior to Posterior: The Complete Bayesian Workflow
Worked Example: Criminal-Specific Investigation
Consider a criminal investigation where a subject is one of four individuals with access to the scene. No other evidence distinguishes among them. The examiner conducts a validated polygraph examination using a comparison question technique.
Prior probability of deception: 25% (1 in 4 suspects is expected to be the perpetrator) Prior odds: 25/75 = 1:3 (or 0.333) Polygraph result: Numerical score of -6 (indicating deception) Likelihood ratio for this score: 12 (based on validated research — this score is 12 times more likely in deceptive subjects)
Posterior odds: 0.333 x 12 = 4:1 Posterior probability of deception: 4/5 = 80%
The polygraph test has shifted the probability of deception from 25% to 80% — a substantial update driven by strong test evidence. However, note that despite a likelihood ratio of 12, the posterior probability does not reach 95% because the prior was only 25%. If the prior had been 50%, the same likelihood ratio would yield approximately 92%.
Worked Example: Pre-Employment Screening
Now consider a pre-employment screening context where 1,000 applicants are tested and the base rate of relevant deception is estimated at 10%.
Prior probability of deception: 10% Prior odds: 10/90 = 1:9 (or 0.111) Polygraph result: Same numerical score of -6 Same likelihood ratio: 12
Posterior odds: 0.111 x 12 = 1.333:1 Posterior probability of deception: 1.333/2.333 = approximately 57%
The identical test evidence — the same score, the same likelihood ratio — yields only a 57% posterior probability of deception in the screening context compared to 80% in the investigation context. This stark difference illustrates why base rates and prior probabilities are essential components of accurate polygraph interpretation, not optional considerations.
Nelson and Turner's Bayesian analysis further demonstrated that when the prior probability of deception is 10%, the posterior probability of truth-telling for negative test results approaches 99% [4]Verified Bayesian Probabilities of Deception and Truth-Telling for Single and Repeated Polygraph Examinations
Confirms Nelson and Turner's research on Bayesian posterior probabilities for single and repeated polygraph exams, including results for various prior probability distributions.. This means that in low base rate environments, a passed polygraph provides extremely high confidence in truthfulness — a powerful finding that underscores the test's value in screening contexts.
For information on how these statistical principles interact with reporting standards, see our article on APA forensic standards for polygraph examiners.
Frequentist vs. Bayesian Methods in Polygraph Analysis
The Frequentist Approach
Frequentist statistics, the dominant paradigm in most scientific fields for much of the 20th century, focuses on the properties of statistical procedures over many repetitions. In polygraph testing, a frequentist analysis asks: 'What is the probability of obtaining this test result if the subject is truthful (or deceptive)?'
Key frequentist concepts in polygraph testing include sensitivity (the proportion of deceptive subjects correctly identified), specificity (the proportion of truthful subjects correctly identified), and confidence intervals. The APA meta-analysis reported these frequentist metrics: event-specific diagnostic accuracy of 89% with a confidence interval of 83-95%, and overall accuracy across all validated techniques of 87% (CI: 80-94%) [7]Verified Meta-Analytic Survey of Criterion Accuracy of Validated Polygraph Techniques
Confirms 38 studies, 3,723 examinations, 11,737 scored results. Event-specific accuracy: 89% (CI: 83-95%). Multi-issue screening: 85% (CI: 77-93%). Overall: 87% (CI: 80-94%)..
P-values represent another frequentist tool used in polygraph research. For a detailed exploration of how p-values function in polygraph testing, see our guide on P-values in polygraph testing.
The Bayesian Advantage
Bayesian methods offer several advantages for polygraph interpretation. Rather than asking about the probability of the data given a hypothesis (the frequentist question), Bayesian analysis directly answers the question examiners actually care about: 'Given this test result, what is the probability that the subject is deceptive?'
Nelson's work in the APA Magazine demonstrated that Bayesian analysis provides a natural framework for incorporating contextual information through the prior probability [3]Verified Base Rates and Prior Probabilities in Polygraph Testing — APA Magazine 2016, Vol. 49, Issue 1
Confirms Nelson's 2016 article on base rates, prior probabilities, and their interaction with test accuracy in the APA Magazine January/February 2016 issue.. The Bayesian approach also handles sequential testing naturally — when an examinee undergoes repeated polygraph examinations, the posterior probability from the first test becomes the prior for the second, allowing evidence to accumulate mathematically [4]Verified Bayesian Probabilities of Deception and Truth-Telling for Single and Repeated Polygraph Examinations
Confirms Nelson and Turner's research on Bayesian posterior probabilities for single and repeated polygraph exams, including results for various prior probability distributions..
APA Standard 1.8.3 states that probabilistic information shall be provided along with categorical test results of evidentiary exams [2]Verified Raymond Nelson Research Profile
Confirms Raymond Nelson is a polygraph researcher who has published extensively on Bayesian analysis, prior probabilities, posterior probabilities, ESS-M scoring, and APA standards including Standard 1.8.3., making Bayesian methods increasingly relevant for modern polygraph practice. A comprehensive comparison of these statistical approaches is available in our dedicated article on Bayesian vs. Frequentist statistics in polygraph testing.
Calculating Prior Probabilities: A Step-by-Step Approach
The Six-Step Process
Calculating prior probabilities for polygraph testing involves several considerations and data sources. Understanding the fundamental steps is essential for any examiner seeking to apply Bayesian methods.
Step 1: Identify the Testing Population. Define the relevant population from which the subject is drawn. Is this pre-employment screening of general applicants? A criminal investigation with a defined suspect pool? A post-conviction monitoring examination? The population determines which base rate data is most applicable.
Step 2: Research Published Base Rates. Consult published research on deception base rates for the relevant testing context. The APA meta-analysis [7]Verified Meta-Analytic Survey of Criterion Accuracy of Validated Polygraph Techniques
Confirms 38 studies, 3,723 examinations, 11,737 scored results. Event-specific accuracy: 89% (CI: 83-95%). Multi-issue screening: 85% (CI: 77-93%). Overall: 87% (CI: 80-94%)., peer-reviewed journals, and validated research studies provide empirical data on base rates in various polygraph applications. When direct data is unavailable, examiners may use analogous populations or conservative estimates.
Step 3: Consider Contextual Information. Evaluate case-specific information that may adjust the prior. In criminal investigations, the strength of corroborating evidence, the number of suspects, and the nature of the allegation all provide information that can refine the prior estimate. However, examiners must be careful not to let subjective bias contaminate this assessment — a concern also explored in our article on factors that affect polygraph accuracy.
Step 4: Select an Appropriate Prior. Based on population data and contextual information, select a prior probability. In the absence of strong prior information, many practitioners default to a 50% prior (maximum uncertainty), which is the most conservative and least biased starting point. This is sometimes called an 'uninformative' or 'flat' prior.
Step 5: Document the Prior Selection Rationale. Record the basis for the selected prior probability. Transparency in prior selection is essential for defensibility. The rationale should be documented in the case file and available for review by quality assurance personnel, attorneys, or courts.
Step 6: Apply Bayesian Updating. After conducting the polygraph test and obtaining numerical scores, apply Bayes' theorem to combine the prior with the test evidence (expressed as a likelihood ratio) to produce the posterior probability.
Choosing Between Informative and Uninformative Priors
The choice of a 50% prior — while common and defensible — is not always optimal. Nelson's research demonstrated that when reliable base rate data is available, using an informative prior can produce more accurate posterior probabilities and better classification decisions [3]Verified Base Rates and Prior Probabilities in Polygraph Testing — APA Magazine 2016, Vol. 49, Issue 1
Confirms Nelson's 2016 article on base rates, prior probabilities, and their interaction with test accuracy in the APA Magazine January/February 2016 issue.[4]Verified Bayesian Probabilities of Deception and Truth-Telling for Single and Repeated Polygraph Examinations
Confirms Nelson and Turner's research on Bayesian posterior probabilities for single and repeated polygraph exams, including results for various prior probability distributions.. The key is that the prior must be justified by empirical evidence, not by the examiner's subjective impression of the individual subject's guilt or innocence.
Bernhardt's 2005 research demonstrated another important dimension of prior information: that stimulation tests with effective feedback increased probable-lie test accuracy to 90% and directed-lie test accuracy to 83% [14]Verified Effects of Prior Demonstrations of Polygraph Accuracy on Outcomes of Probable-Lie and Directed-Lie Polygraph Tests
Confirms stimulation test with effective feedback increased probable-lie test accuracy to 90% and directed-lie test accuracy to 83%.. This suggests that the examiner's preparation process — including demonstrations of polygraph accuracy — can influence the quality of the physiological data that feeds into Bayesian calculations.
For examiners new to these concepts, formal polygraph training programs increasingly incorporate Bayesian methods into their curricula. The career path for polygraph examiners now typically includes statistical literacy as an essential competency. The importance of proper methodology is further reflected in APA forensic standards.
How Prior Probabilities Improve Classification Accuracy
Evidence from Research
The value of incorporating prior probabilities into polygraph interpretation is supported by substantial research evidence. The APA's 2011 meta-analysis — encompassing 38 studies, 32 different samples, 45 experiments, 295 scorers, and 11,737 scored results from 3,723 examinations — provides the most comprehensive accuracy data available for Bayesian calculations [7]Verified Meta-Analytic Survey of Criterion Accuracy of Validated Polygraph Techniques
Confirms 38 studies, 3,723 examinations, 11,737 scored results. Event-specific accuracy: 89% (CI: 83-95%). Multi-issue screening: 85% (CI: 77-93%). Overall: 87% (CI: 80-94%).. Event-specific diagnostic tests achieved 89% accuracy, while the combination of all validated techniques produced 87% accuracy [7]Verified Meta-Analytic Survey of Criterion Accuracy of Validated Polygraph Techniques
Confirms 38 studies, 3,723 examinations, 11,737 scored results. Event-specific accuracy: 89% (CI: 83-95%). Multi-issue screening: 85% (CI: 77-93%). Overall: 87% (CI: 80-94%)..
Abrams' 1989 comprehensive survey of published polygraph literature reported an overall accuracy level of.89 [15]Verified Survey of Published Polygraph Literature on Accuracy
Confirms Abrams' comprehensive survey reported overall accuracy level of.89., while Honts and Peterson's 1997 review reported accuracy exceeding.90 [16]Verified Confidence, Accuracy, and Utility of Polygraph Decisions
Confirms examination of the relationship between examiner confidence and decision accuracy, highlighting that structural improvements reduce error rates., consistent with earlier findings by Raskin and Podlesny (1979) who also reported accuracy exceeding.90 using the CQT format [17]Verified The Discussion of Comparison Questions Between List Repetitions Is Associated with Increased Test Accuracy
Confirms between-chart discussion of comparison questions significantly improved CQT accuracy.. These converging accuracy estimates from independent researchers across decades provide robust data for populating Bayesian models.
Rigdon's 2018 Bayesian analysis provided a formal statistical framework for combining these accuracy estimates with prior probability information to generate posterior distributions [8]Verified Exact Bayesian Inference for Assessing the Accuracy of Polygraph Testing
Confirms Rigdon (2018) developed a formal Bayesian framework for assessing polygraph accuracy using posterior distributions.. The Bayesian approach enabled researchers to obtain not just point estimates of accuracy, but full probability distributions that quantify uncertainty — a significant advancement over traditional frequentist reporting.
The Role of Examiner Competence
Bayesian methods enhance but cannot replace examiner competence. Kleinmuntz and Szucko's 1984 research examined the relationship between examiner confidence and decision accuracy, highlighting that structural improvements to testing protocols can reduce error rates [18]Verified The Accuracy of Physiological Detection of Deception for Subjects with Prior Knowledge
Foundational research relevant to understanding how prior knowledge affects physiological detection accuracy.. Honts' 1999 study demonstrated that between-chart discussion of comparison questions significantly improved CQT accuracy [19]Verified Mental and Physical Countermeasures Reduce the Accuracy of the Concealed Knowledge Test
Confirms countermeasures can reduce detection accuracy, relevant to understanding factors affecting likelihood ratio reliability., showing that procedural refinements feed directly into the quality of data used in Bayesian calculations.
Research by Rovner (1986) on the accuracy of physiological detection for subjects with prior knowledge [20]Verified Bayesian Inference for Interpretation of Polygraph Results in the Courtroom
Confirms that polygraphers rarely describe positive and negative predictive value; presents Bayesian analysis of polygraph data for courtroom use. and by Honts et al. (1996) on the impact of countermeasures further illustrates that the quality of physiological data — and therefore the reliability of likelihood ratios used in Bayesian updating — depends on proper test administration. These findings underscore why stimulation tests and acquaintance tests play an important role in establishing test validity.
The PEACE model and cognitive interviewing techniques can complement Bayesian polygraph analysis by improving the quality of the pre-test interview phase, which in turn affects the reliability of the physiological data collected during testing.
Practical Applications for Polygraph Examiners
Reporting Probabilistic Information
APA Standard 1.8.3 requires that probabilistic information be provided along with categorical test results of evidentiary exams [2]Verified Raymond Nelson Research Profile
Confirms Raymond Nelson is a polygraph researcher who has published extensively on Bayesian analysis, prior probabilities, posterior probabilities, ESS-M scoring, and APA standards including Standard 1.8.3.. This standard directly supports the use of Bayesian methods in professional practice. Examiners conducting evidentiary examinations should be prepared to report not just 'Deception Indicated' or 'No Deception Indicated,' but the statistical confidence associated with those determinations.
Nelson's work on how to write probability information in evidentiary polygraph reports [2]Verified Raymond Nelson Research Profile
Confirms Raymond Nelson is a polygraph researcher who has published extensively on Bayesian analysis, prior probabilities, posterior probabilities, ESS-M scoring, and APA standards including Standard 1.8.3. provides practical guidance for translating Bayesian calculations into clear, defensible report language. The ESS-M (Empirical Scoring System - Multinomial) represents one of the modern scoring systems that incorporates Bayesian classification, using a multinomial combinatoric distribution calculated under the basic analytic theory of the polygraph test [2]Verified Raymond Nelson Research Profile
Confirms Raymond Nelson is a polygraph researcher who has published extensively on Bayesian analysis, prior probabilities, posterior probabilities, ESS-M scoring, and APA standards including Standard 1.8.3..
Examiners should be cautious about red flags in polygraph advertising that claim unrealistic accuracy figures without statistical context. Similarly, understanding Bayesian methods helps examiners identify fake polygraph results by recognizing when reported confidence levels are inconsistent with the testing context.
Sequential Testing and Retesting Strategies
One of the most powerful applications of Bayesian analysis in polygraph testing is its ability to handle sequential examinations. Nelson and Turner demonstrated that when an examinee undergoes repeated polygraph tests, the posterior probability from the first examination can serve as the prior for the second [4]Verified Bayesian Probabilities of Deception and Truth-Telling for Single and Repeated Polygraph Examinations
Confirms Nelson and Turner's research on Bayesian posterior probabilities for single and repeated polygraph exams, including results for various prior probability distributions..
In their analysis, when the prior probability of deception was initially set at 50% and the first examination produced a deceptive result, the posterior probability of deception increased to 84.9% [4]Verified Bayesian Probabilities of Deception and Truth-Telling for Single and Repeated Polygraph Examinations
Confirms Nelson and Turner's research on Bayesian posterior probabilities for single and repeated polygraph exams, including results for various prior probability distributions.. When this elevated posterior was used as the prior for a second examination that also produced a deceptive result, confidence increased further. Similarly, truthful results on sequential tests compound to produce increasingly high confidence in truthfulness [4]Verified Bayesian Probabilities of Deception and Truth-Telling for Single and Repeated Polygraph Examinations
Confirms Nelson and Turner's research on Bayesian posterior probabilities for single and repeated polygraph exams, including results for various prior probability distributions..
However, their analysis also showed important caveats: no increase in decision accuracy can be expected when conducting two examinations unless the test results concur [4]Verified Bayesian Probabilities of Deception and Truth-Telling for Single and Repeated Polygraph Examinations
Confirms Nelson and Turner's research on Bayesian posterior probabilities for single and repeated polygraph exams, including results for various prior probability distributions.. This finding has practical implications for retesting policies and for how examiners interpret discordant results across multiple examinations.
Challenges and Limitations of Prior Probabilities
Data Availability and Quality
The primary challenge in applying Bayesian methods to polygraph testing is obtaining reliable base rate data. While the APA meta-analysis provides robust accuracy figures for well-conducted examinations [7]Verified Meta-Analytic Survey of Criterion Accuracy of Validated Polygraph Techniques
Confirms 38 studies, 3,723 examinations, 11,737 scored results. Event-specific accuracy: 89% (CI: 83-95%). Multi-issue screening: 85% (CI: 77-93%). Overall: 87% (CI: 80-94%)., empirical data on deception base rates for specific testing populations is limited. For many contexts — particularly private testing and specialized screening applications — direct base rate data may not exist.
The Oxford University Press journal 'Law, Probability and Risk' published research noting that professional polygraphers rarely describe the positive and negative predictive value of the test, which are the Bayesian metrics most directly relevant to individual case interpretation. This gap between available accuracy data and the predictive value information needed for individual cases highlights the practical challenges of implementing Bayesian methods.
Additionally, certain populations present unique challenges. Testing elderly examinees or individuals with conditions such as Parkinson's disease may require adjustments to both testing protocols and statistical models.
Subjectivity in Prior Selection
A common criticism of Bayesian analysis is that the selection of prior probabilities can introduce subjectivity. If two examiners choose different priors for the same subject, they will arrive at different posterior probabilities even when the test data is identical. This concern is legitimate but manageable through several safeguards.
First, using a default 50% prior when no reliable base rate data is available ensures maximum objectivity. Second, requiring documentation of the rationale for any non-default prior creates transparency and accountability. Third, sensitivity analysis — calculating posterior probabilities across a range of plausible priors — demonstrates how robust conclusions are to changes in the prior assumption.
Nelson's work demonstrated results across a distribution of prior odds from 1 in 10 to 9 in 10 [4]Verified Bayesian Probabilities of Deception and Truth-Telling for Single and Repeated Polygraph Examinations
Confirms Nelson and Turner's research on Bayesian posterior probabilities for single and repeated polygraph exams, including results for various prior probability distributions., showing how posterior probabilities vary systematically with the prior. This type of sensitivity analysis allows examiners to present a range of conclusions rather than a single point estimate, providing a more complete picture of the evidence.
It is also important that emerging AI-based analysis tools are evaluated within this Bayesian framework. Our analysis of whether AI can outperform human polygraph examiners explores how automated systems handle probabilistic reasoning.
Identify the Testing Population
Define the relevant population: pre-employment screening, criminal investigation, PCSOT, or private testing. The population determines which base rate data applies.
Research Published Base Rates
Consult the APA meta-analysis, peer-reviewed journals, and validated studies for empirical base rate data. When direct data is unavailable, use conservative estimates or analogous populations.
Evaluate Contextual Information
Consider case-specific factors that may adjust the prior: strength of corroborating evidence, number of suspects, nature of the allegation. Avoid letting subjective bias contaminate the assessment.
Select and Document the Prior
Choose a prior probability justified by empirical evidence. Default to 50% when no reliable data exists. Document the rationale for transparency and defensibility.
Conduct the Examination and Calculate Likelihood Ratio
Administer the validated polygraph test. Obtain numerical scores and determine the corresponding likelihood ratio from validated research distributions.
Compute the Posterior Probability
Apply Bayes' theorem: multiply the prior odds by the likelihood ratio to obtain posterior odds, then convert to a posterior probability. Report this alongside the categorical decision.
Pros
- Transforms subjective pass/fail judgments into quantified probabilistic assessments backed by mathematical rigor
- Reduces false positive rates in low base rate screening environments by properly accounting for population statistics
- Provides a natural framework for combining evidence from sequential polygraph examinations
- Enables transparent documentation of decision rationale through explicit prior selection and calculation
- Aligns with APA Standard 1.8.3 requirements for probabilistic reporting in evidentiary examinations
- Improves communication of results to attorneys, courts, and decision-makers by expressing confidence levels
- Allows sensitivity analysis across a range of priors, demonstrating robustness of conclusions
Cons
- Requires reliable base rate data that may not be available for all testing contexts and populations
- Introduces potential subjectivity in prior selection if documentation and justification standards are not maintained
- Demands statistical literacy that not all practicing examiners currently possess
- Cannot compensate for poor test administration or compromised physiological data quality
- May be misunderstood by non-technical audiences if not clearly explained in reports and testimony
Frequently Asked Questions
What is a prior probability in polygraph testing?
A prior probability is the statistical estimate of the likelihood that a polygraph subject is deceptive or truthful before any test data is collected. It is based on the base rate of deception in the population being tested and the specific context of the examination — not on the examiner's subjective impression of the individual. For example, in a criminal investigation with four equally likely suspects, the prior probability of deception for each is 25%.
Why do base rates matter so much in polygraph interpretation?
Base rates determine the predictive value of test results. A 90% accurate polygraph test yields a 90% positive predictive value when the base rate of deception is 50%, but only about 32% when the base rate drops to 5%. This means the same test result can indicate very different probabilities of deception depending on the testing context. Ignoring base rates leads to the base rate fallacy — a common statistical error that can cause examiners to substantially overestimate or underestimate the probability of deception.
What prior probability should I use if I have no base rate data?
When reliable base rate data is unavailable, many practitioners default to a 50% prior probability, also known as an 'uninformative' or 'flat' prior. This represents maximum uncertainty and is the most conservative, least biased starting point. It means the examiner is assigning equal probability to deception and truthfulness before the test. While not always optimal, a 50% prior ensures that the test result alone drives the conclusion, which is the most defensible position when empirical base rate data is lacking.
How does Bayes' theorem improve polygraph accuracy?
Bayes' theorem does not change the raw accuracy of the polygraph instrument itself. Rather, it improves the accuracy of the conclusions drawn from test results by properly incorporating base rate information. When examiners combine the test's likelihood ratio with an appropriate prior probability, the resulting posterior probability more accurately reflects the true probability of deception. The APA meta-analysis found 89% accuracy for event-specific tests [7]Verified Meta-Analytic Survey of Criterion Accuracy of Validated Polygraph Techniques
Confirms 38 studies, 3,723 examinations, 11,737 scored results. Event-specific accuracy: 89% (CI: 83-95%). Multi-issue screening: 85% (CI: 77-93%). Overall: 87% (CI: 80-94%)., and applying Bayesian methods to these results with properly calibrated priors produces more reliable and defensible classification decisions.
What is a likelihood ratio in polygraph testing?
A likelihood ratio compares the probability of observing a particular test score if the subject is deceptive versus if the subject is truthful. For example, a likelihood ratio of 12 means the observed score is 12 times more likely in a deceptive person than in a truthful one. Different numerical scores on a polygraph correspond to different likelihood ratios — extreme scores have high likelihood ratios providing strong evidence, while scores near the decision threshold have likelihood ratios close to 1, providing little diagnostic information.
Can Bayesian analysis be applied to repeated polygraph tests?
Yes, and this is one of the most powerful applications of Bayesian methods in polygraph practice. When an examinee undergoes multiple tests regarding the same issue, the posterior probability from the first test becomes the prior for the second. Nelson and Turner's research showed that when two consecutive tests both indicate deception, starting from a 50% prior, the confidence in deception increases substantially — from 84.9% after the first test to even higher levels after the second [4]Verified Bayesian Probabilities of Deception and Truth-Telling for Single and Repeated Polygraph Examinations
Confirms Nelson and Turner's research on Bayesian posterior probabilities for single and repeated polygraph exams, including results for various prior probability distributions.. However, this only works when test results concur; discordant results do not increase accuracy.
How does the Bayesian approach differ from traditional polygraph scoring?
Traditional polygraph scoring typically uses a frequentist approach: the examiner assigns numerical scores to chart features and compares total scores against predetermined cut-off values to reach a categorical decision (DI, NDI, or Inconclusive). Bayesian analysis adds a layer of contextual reasoning by incorporating the base rate of deception in the relevant population. The categorical decision is supplemented with a posterior probability that quantifies confidence. APA Standard 1.8.3 now requires probabilistic information for evidentiary exams, reflecting the profession's recognition of the Bayesian framework's value.
What is the base rate fallacy and why is it dangerous in polygraph work?
The base rate fallacy occurs when decision-makers focus exclusively on test accuracy while ignoring the prevalence of the condition being tested for. In polygraph work, this means interpreting a 'Deception Indicated' result as highly reliable without considering how common deception is in the testing population. The NRC (2003) demonstrated that in security screening with very low base rates, even a highly accurate test would produce far more false positives than true positives [6]Verified The Polygraph and Lie Detection — National Research Council
Confirms the NRC 2003 report's analysis of base rate effects on polygraph screening, including the finding that low base rates produce high false positive rates even with accurate tests.. Examiners who understand the base rate fallacy can properly communicate the actual confidence in their findings.
Sources & References
Confirms Thomas Bayes was born c. 1701, was an English statistician, philosopher, and Presbyterian minister. His work was published posthumously in 1763 by Richard Price.
Confirms Raymond Nelson is a polygraph researcher who has published extensively on Bayesian analysis, prior probabilities, posterior probabilities, ESS-M scoring, and APA standards including Standard 1.8.3.
Confirms Nelson's 2016 article on base rates, prior probabilities, and their interaction with test accuracy in the APA Magazine January/February 2016 issue.
Confirms Nelson and Turner's research on Bayesian posterior probabilities for single and repeated polygraph exams, including results for various prior probability distributions.
Confirms Nelson (2016) published the scientific analytic theory of polygraph testing in APA Magazine 49(5), 69-82.
Confirms the NRC 2003 report's analysis of base rate effects on polygraph screening, including the finding that low base rates produce high false positive rates even with accurate tests.
Confirms 38 studies, 3,723 examinations, 11,737 scored results. Event-specific accuracy: 89% (CI: 83-95%). Multi-issue screening: 85% (CI: 77-93%). Overall: 87% (CI: 80-94%).
Confirms Rigdon (2018) developed a formal Bayesian framework for assessing polygraph accuracy using posterior distributions.
Confirms the OTA's analysis of base rate effects in screening contexts, including the example of 1 guilty person per 1,000 screened and resulting false positive rates.
Confirms base rate problem in pre-employment screening, inconclusive result handling, and Information Gain analysis showing peak information value at base rate of 19%.
Confirms Honts and Peterson reported polygraph accuracy exceeding.90, consistent with other major accuracy reviews.
Confirms Raskin and Podlesny reported polygraph accuracy exceeding.90 using the CQT format.
Confirms Honts and Schweinle's Information Gain analysis adapted for CQT and the Bayesian-based approach to describing base rate impact on polygraph outcomes.
Confirms stimulation test with effective feedback increased probable-lie test accuracy to 90% and directed-lie test accuracy to 83%.
Confirms Abrams' comprehensive survey reported overall accuracy level of.89.
Confirms examination of the relationship between examiner confidence and decision accuracy, highlighting that structural improvements reduce error rates.
Confirms between-chart discussion of comparison questions significantly improved CQT accuracy.
Foundational research relevant to understanding how prior knowledge affects physiological detection accuracy.
Confirms countermeasures can reduce detection accuracy, relevant to understanding factors affecting likelihood ratio reliability.
Confirms that polygraphers rarely describe positive and negative predictive value; presents Bayesian analysis of polygraph data for courtroom use.
Bayesian reasoning sharpens interpretation of results, so for an expertly conducted exam, find a lie detector test near you and check pricing.