Polygraph accuracy is one of the most searched — and most misunderstood — questions about lie-detection. The short version: a properly administered, single-issue polygraph is correct in roughly nine cases out of ten, far better than unaided human judgment but well short of certainty. This page sets out the published accuracy rates by test type, explains what “accuracy” actually measures, and shows why the same number can mean very different things in different situations. Full citations are in Sources.

Key points

  • Single-issue (event-specific) tests: about 89% accurate (APA 2011)[1].
  • At the top end, higher still: the best-validated single-issue techniques individually exceed 90%, and well-controlled studies report decision accuracy in the mid-90s — see studies that report higher accuracy.
  • Multiple-issue / screening tests: about 85%, and far more error-prone at low base rates[2].
  • Recent comprehensive meta-analysis: AUC about .91 for the Comparison Question Test">Comparison Question Test[3].
  • Unaided humans: only about 54% — barely better than a coin flip[4]. This is the benchmark the polygraph beats.
  • No test is perfect: a polygraph result is evidence to be weighed, never a verdict — and it should never be the sole basis for a decision.

Contents

1. The short answer: how accurate is a polygraph?

When a competent examiner runs a single-issue, event-specific examination — one named person, one known allegation — the published accuracy is around 89%, with the most comprehensive recent meta-analysis putting the discrimination of the Comparison Question Test at an AUC of about .91[5]. That is high — far above the roughly 54% that trained humans manage unaided — but it is not certainty, and it is not the same across every type of test.

The one thing to remember

There is no single “polygraph accuracy” number. Accuracy depends on which test is run (single-issue vs screening), how it is scored and administered, and the base rate of guilt in the situation. A headline percentage with none of that context is close to meaningless.

2. Polygraph accuracy rates by test type

The single most common mistake in discussions of polygraph accuracy is treating all tests as one. They are not. Here are the published figures, excluding inconclusive results:

Test type Used for Accuracy Source
Single-issue (event-specific) One named suspect, one known allegation ~89% (.83–.95) APA 2011[6]
Multiple-issue / screening No specific allegation (e.g. vetting, periodic testing) ~85% (.77–.93) APA 2011[7]
Comparison Question Test (recent meta) The dominant field deception test">deception test AUC ~.91 (rdec .69) Honts 2021[8]
Concealed Information Test Recognition of undisclosed crime details AUC up to ~.95 Ben-Shakhar & Elaad 2003[9]
Unaided human judgment (baseline) No instrument ~54% Bond & DePaulo 2006[10]

Field beats laboratory

A forty-year worry was that high accuracy came only from artificial laboratory experiments. The most recent meta-analysis settled it the other way: real-world field studies returned higher discrimination (rdec .76) than laboratory studies (.64)[11]. For the full evidence review, see Does the Polygraph Work?

3. What “accuracy” actually measures

“Accuracy” bundles together two different things, and the polygraph is better at one than the other:

Sensitivity (catching the guilty)
How often a genuinely deceptive person is correctly flagged. The polygraph’s sensitivity is strong — deceptive calls in the recent meta-analysis were about 91.6% correct[12].
Specificity (clearing the innocent)
How often a genuinely truthful person is correctly cleared. This is the weaker side: truthful calls were about 78.9% correct, and the published evidence for specificity is thinner than for sensitivity[13].

It is also worth separating reliability (would two examiners score the same charts the same way? — generally yes) from validity (does the call match the truth?). When people argue about polygraph accuracy, they almost always mean validity. If you are worried about a specific wrong result, see Can a Polygraph Be Wrong? Accuracy, Errors & Limits.

4. Why the same result can mean different things

Here is the counter-intuitive part. An 89%-accurate test does not mean any single result is 89% likely to be right. That depends on the base rate — how likely guilt was before the test. At a very low base rate (say, screening a large population where almost everyone is innocent), even a good test produces many false alarms, simply because there are so many more innocent people to misclassify.

Direction + base rate = how much weight

A failed test where guilt was already likely, and a passed test where guilt was already unlikely, are both highly informative. The same results in the opposite settings tell you much less. This is exactly why screening at low base rates is treacherous[14] while single-issue testing of a real suspect is not — and why a polygraph result should be weighed alongside other evidence, never treated as a verdict.

5. What reduces polygraph accuracy in practice

  • Test type. Multiple-issue screening is inherently lower-accuracy than single-issue testing (see the table above).
  • Examiner quality and scoring. Real-world accuracy depends on competent administration and validated, often algorithmic, scoring. Regulation is uneven, so examiner choice matters — see How to Choose a Polygraph Examiner.
  • Decision policy. How an agency sets its cut-offs (and how it treats inconclusive results) changes how many innocent people are wrongly flagged, independent of the instrument.
  • Countermeasures. Spontaneous attempts are largely ineffective and detectable; trained, coached countermeasures are a more serious threat. See Can You Beat a Polygraph?

6. Polygraph accuracy versus the alternative

The fair comparison is not “polygraph versus perfect.” It is “polygraph versus the judgment that would otherwise be made.” Decades of research show that people — including trained police officers and other professionals — detect deception at about 54%, barely above chance, and that experience does not improve this[15]. Measured against that benchmark, a validated polygraph adds substantial information to a decision — which is the practical case for using it at all.

7. Studies that report higher accuracy

The ~89% headline is an aggregate that pools many techniques and studies. Several credible lines of evidence report figures above it:

  • Well-controlled mock-crime studies. The classic meta-analysis of mock-crime experiments found correct-decision rates ranging as high as 100%, with the better-designed studies clustering well above 90%[16].
  • Specific validated techniques. In the APA 2011 survey, five techniques individually exceeded 90% mean accuracy with low inconclusive rates — among them the Zone Comparison Test">Zone Comparison Test">Utah Zone Comparison Test and the Federal You-Phase, with the Zone Comparison Test scored by the Empirical Scoring System">Empirical Scoring System at about 92%[17]. The .89 aggregate understates the best-performing validated methods.
  • Algorithmic scoring. Computer scoring systems such as OSS-3 report accuracy across roughly the 85–100% range depending on the test format, with essentially perfect score reliability[18].

Read the very highest numbers with care

Two techniques report near-perfect accuracy: the Integrated Zone Comparison Technique (IZCT) at essentially 100% when inconclusives are excluded[19], and the Matte Quadri-Track Zone Comparison Technique at ~98–100% in a 140-case field study[20]. But the APA 2011 committee treated both as statistical outliers, because their validation was largely conducted by the techniques’ own developers — and a study in which the inventor is also the investigator is not an independent test. The defensible reading: a competently run single-issue test sits in the high-80s to low-90s; figures of 98–100% should prompt the question, “who ran the study?”

8. The bottom line

So, how accurate is a polygraph? For a properly run single-issue test, around 89%, with strong discrimination (AUC ~.91) confirmed in field conditions — clearly better at catching the guilty than at clearing the innocent, clearly better than any unaided human judgment, and clearly lower for multiple-issue screening. It is highly accurate but never perfect. The honest reading is that a polygraph result is strong evidence to be weighed in context, not a verdict to be pronounced. For the full peer-reviewed picture, read Does the Polygraph Work? A Field-Grade Reading of the Evidence.

9. Frequently asked questions

How accurate is a polygraph test?

A properly administered single-issue (event-specific) polygraph is about 89% accurate, and the most comprehensive recent meta-analysis reports an AUC of about .91 for the Comparison Question Test. Multiple-issue screening tests are lower, around 85%, and less reliable at low base rates of guilt.

Is the polygraph more accurate than a human?

Yes, substantially. Unaided humans — including trained professionals — detect deception at only about 54%, barely above chance. A validated polygraph is far more accurate, which is the main practical argument for using it.

What is the polygraph’s false-positive rate?

The polygraph is better at catching the guilty than at clearing the innocent, so false positives (a truthful person failing) are its weaker area. The exact rate depends on the test type, the scoring threshold, and the base rate of guilt. A result should never be the sole basis for a decision.

Does a high accuracy rate mean my result is definitely right?

No. An 89% accurate test does not mean a single result is 89% likely to be correct — that depends on how likely guilt was beforehand (the base rate). This is why a polygraph result should be weighed alongside other evidence rather than treated as proof.

10. Sources

Each inline footnote links to its entry here; hover or tap a footnote for the underlying fact and proof. For the full reference list and detailed argument, see Does the Polygraph Work?

  1. American Polygraph Association (2011). Meta-analytic survey of criterion accuracy of validated polygraph techniques. Polygraph, 40(4), 194–305.
  2. Honts, C. R., Thurber, S. & Handler, M. (2021). A comprehensive meta-analysis of the comparison question polygraph test. Applied Cognitive Psychology, 35(2), 411–427.
  3. Bond, C. F. & DePaulo, B. M. (2006). Accuracy of deception judgments. Personality and Social Psychology Review, 10(3), 214–234. See also Hartwig & Bond (2011), Psychological Bulletin, 137(4), 643–659.
  4. Ben-Shakhar, G. & Elaad, E. (2003). The validity of psychophysiological detection of information with the Guilty Knowledge Test: A meta-analytic review. Journal of Applied Psychology, 88(1), 131–151.
  5. National Research Council (2003). The Polygraph and Lie Detection. Washington, DC: National Academies Press.
  6. Kircher, J. C., Horowitz, S. W. & Raskin, D. C. (1988). Meta-analysis of mock-crime studies of the control question polygraph technique. Law and Human Behavior, 12(1), 79–90.
  7. Nelson, R., Krapohl, D. & Handler, M. (2008). Brute force comparison: A Monte Carlo study of the Objective Scoring System version 3 (OSS-3) and human polygraph scorers. Polygraph, 37(3), 185–215.
  8. Gordon, N. J., Mohamed, F. B., Faro, S. H., Platek, S. M., Ahmad, H. & Williams, J. M. (2006). Integrated zone comparison polygraph technique accuracy with scoring algorithms. Physiology & Behavior, 87(2), 251–254.
  9. Mangan, D. J., Armitage, T. E. & Adams, G. C. (2008). A field study on the validity of the Quadri-Track Zone Comparison Technique. Physiology & Behavior, 95(1–2), 17–23.

This page was prepared for the British Polygraph Society as a plain-English summary of published polygraph accuracy figures. It is not a substitute for the primary research; readers making professional or legal use of the material should consult the original sources and a qualified examiner.

↑ Top

← All News