Knowledgebase
Question Last Updated on: 2/6/2025
Refers to the type of numerical scale used to obtain the scores assigned to the Relevant Questions. Manual scores can be thought of as either multinomial or Likert type.
ESS scores are 3 position multinomial scores with weighted EDA values. ESS scores are multinomial because each score can have one of three possible values [-, 0, +]. Scores are assigned to each analysis spot, consisting of a Relevant Question (RQ) and a Comparison Question (CQ), depending on whether the change in physiological activity is objectively greater at the RQ or CQ, or are objectively equivalent. In contrast, when there are two possible outcome values (e.g., [-, +]) these scores would be binomial. ESS scores are weighed because EDA scores can have the integer values [-2, 0, +2] while other sensors are limited to the integer values [-1, 0, +1]. ESS scores are aggregated through summation. A result of these weighted scores is that EDA data contribute more information to test scores and test results. ESS scores are assigned using only the primary scoring features: electrodermal amplitude of increase in phasic activity, cardiovascular amplitude of increase in relative blood pressure, suppression or reduction of respiratory activity, and suppression or reduction of vasomotor activity. ESS scoring features are referred to as Kircher features and have the advantage that they can be quantified objectively.
Three-position scores are unweighted integer scores [-1, 0, +1] for all polygraph recording sensors. These scores indicate whether a greater change in physiology is observed in response to the RQ or CQ within an analysis spot. Three position scores can be multinomial when the difference is based on objective data using primary response features, referred to as Kircher features. Three position scores can also be obtained as Likert scores, such as when using the Federal 3-Position method which involves a combination of primary and secondary features. In contrast to the objective quantification of primary response features, secondary features can involve pattern recognition methods that are less easily subject to quantification, and which therefor involve a greater degree of subjectivity - along with greater potential for inter-scorer variability. Although superficially like multinomial scores, Likert scores were devised to obtain numerical values for subjective information.
Seven-position scores are unweighted integer scores [-3, -2, -1, 0, +1, +2, +3] for all polygraph recording sensors. Seven-position scores attempt to capture more information than three-position score by representing the degree of difference in addition to whether the greater change in physiological activity was observed at the RQ or CQ within each analysis spot. Seven-position scores are Likert type values representative of subjective judgments - with no theoretical basis (subject to mathematical and logical proof) for the locations of boundary or threshold differences between the seven-position integer scores [-3, -2, -1, 0, +1, +2, +3], despite the formulation of arbitrary or empirical rules and procedures to define these differences. Several different seven-position scoring methods have been described over many decades. Emphasis on evidence-based practice has resulted in obvious practical and empirical similarities between some seven-position scoring methods - especially those published by ASTM, the researchers at the University of Utah, and U.S. Federal polygraph programs. Seven-position scores are obtained using a combination of primary and secondary scoring features. In contrast to the objective quantification of primary response features, secondary features can involve pattern recognition methods that are less easily subject to quantification, and which therefor involve a greater degree of subjectivity - along with greater potential for inter-scorer variability. Seven-position scores may also be subject to inter-scorer variability associated with the application of different boundary or threshold procedures for the different integer values. In actual field practice the entire range of seven-position scores may be used only very rarely for some sensors (e.g., respiration scores are often constrained to the range [-1, 0, +1]), and this may contribute to some reduction of inter-scorer variability.
Numerical cutscores are the information thresholds used to understand the strength and meaning of grand total and subtotal scores, and to classify polygraph test results as either deceptive or truthful.
Statistical cutscores are obtained through the computation of a statistical reference model. This model might be a simple table of all possible score values, and the probabilities associated with each, and might also take the form of a mathematical/statistical function to compute a reference table. ESS cutscores were obtained using both empirical and mathematical/theoretical methods. Statistical reference models are used to select cutscores to optimize desired effect sizes that relate to organizational goals and mission objectives. Statistical cutscores are selected to provide a basis for discussion of the probabilistic strength and meaning of categorical polygraph test results. Statistical cutscores also include the use of statistical corrections to reduce the effects of multiplicity error when using subtotal scores. ESS cutscores are selected to attain posterior information that exceeds the strength of prior information at a required level of statistical significance. Published statistical reference models are available for ESS, 3-Position, and 7-Position scores.
Traditional cutscores do not include the use of statistical corrections, and were selected using heuristic methods that do not involve the use of empirical or statistical reference distributions. These cutscores are associated with each different test format and are traditionally used with 3-Position and 7-Position scoring methods.
Prior odds describe the probabilistic strength of information prior to conducting a polygraph test. The strength of information can is often described in the form of the odds or chances that an examinee is actually deceptive. Odds have the advantages that they express probabilistic information using whole numbers and explicitly convey that the information describes the strength of a possible outcome compared to the strength of some other possible outcome. Prior information is then conditioned on the probabilistic strength of the information from the test to obtain the posterior odds. Algebraically, because the prior odds convey the strength of information indicative of deception and truth-telling, posterior information can also be described as pertaining to deception and truth-telling. The default prior is 1 to 1 (.5), indicating that the strength of prior information for the deception and truth-telling is objectively equal. The prior can be changed based on available information.
Alpha indicates the strength of the probabilistic confidence that the posterior odds are actually stronger than the prior odds, while accounting for expected variation (given a small data set of test scores and the possibility that these scores would vary if the test were repeated). Alpha values are used to select ESS (statistical) cutscores, with the default values at .05 for all outcome classifications (DI, NDI, SR, NSR). ESS alphas indicate the (1-alpha x 100%) range of expected variation that is expected to be observed on repeated tested. This is referred to as a Bayesian credible interval (allegorical to a frequentist confidence interval). Cutscores are selected to ensure that the lower limit of the credible interval for posterior odds exceeds the prior odds. In other words, with alpha = .05 there is a 95% probability that the posterior information is stronger than the prior - that the posterior information would continue to exceed the prior with more data (upon repeated testing). Traditional cutscores are not associated with alpha values.
The cut ratio is an added level of control when selecting ESS cutscores. The lower limit of the credible interval for the posterior odds must exceed this value. The default cut ratio is 1 to 1, meaning that ESS cutscores will not permit a classification of deception or truth-telling when the lower limit is less than this value.
The Test of Proportions is a common statistical method used to determine if there are statistically significant imbalances observed proportions when comparing two groups. It is used polygraph data analysis to determine whether observed artifacts (occurring at CQs and RQs) are consistent with a random distribution. When the statistical probability that the artifact pattern is random is very low (exceeds an alpha threshold) it the data can be classified as statistically significant for non-random artifacts, indicating a possibility of strategic or systematic faking during testing. Test results are inconclusive when the Test of Proportion is statistically significant. Alpha for the Test of Proportion is set to .05. This option enables or disables the Test of Proportions.
This option permits the reporting of DI/SR classifications when the Test of Proportions is statistically significant if the test data (test scores) support this classification.
Alpha for the Test of Proportions is set by default to .05
Allows for pneumo scores to be combined into one score as most examiners do in hand scoring.
This setting will combine upper and lower pneumo scores for one value. For example, if the upper pneumo is a +1 and the lower pneumo is a -1, the score added to the RQ subtotal (and grand total) will be 0.
Unable find an answer to the question that you looking for in our FAQ section? Submit your question directly to our customer service representatives.
Join our email list for up-to-date product news Sign Me Up!
3700 Sagamore Pkwy N
Lafayette, IN 47904
Phone: (765) 423-1505
Fax: (765) 423-4111
info@lafayetteinstrument.com
Copyright © 2009 - 2025 Lafayette Instrument Company. All Rights Reserved.
