A Polygraph Technique for Evidentiary Applications
Donald J. Krapohl and Yazmín Vélez
There are three principal testing formats used in polygraphy today. The most common method is the Comparison Question Technique (CQT), which takes on various forms and compositions: Reid, Arther, Backster, DoDPI, Marcy, Gordon, and others. The CQT is the method of choice almost everywhere the polygraph is used in criminal investigations, except in Japan and China. Its popularity stems from its wide range of applications, and the advent of scoring systems to which the CQT is amenable. To a lesser extent, the Relevant-Irrelevant (RI) technique is used, primarily in multiple-issue screening applications. For many years the RI was applied to criminal testing, but it began to decline with the spread of the CQT in the 1950s. The third format can be generically called the Concealed Knowledge Technique (CKT). The family of CKTs includes the Peak of Tension (POT) tests, stimulation tests, and the Guilty Knowledge Test (GKT). While most polygraph practitioners have used POT and stimulation tests, the GKT is less familiar, and it is the GKT that is the focus of this paper.
In 1959, Dr. David T. Lykken reported a testing and scoring system for the electrodermal channel that could be used in criminal investigations. Dr. Lykken’s method, which he called the Guilty Knowledge Test, capitalized on the higher physiological arousal that occurs when a guilty subject recognized a crime-related test item placed on a test among several topically similar but unrelated items. Innocent subjects, those who do not know which of the items are relevant and which are neutral, are less likely to produce electrodermal responses (EDRs) to the crime-related test item.
A test for guilty knowledge was hardly new to the field of lie detection, and the earliest work on guilty knowledge for criminal applications can be traced back at least to the first part of the last century (Crane, 1915). Polygraph examiners may recognize the GKT as another form of POT, to which it is very closely related and has been used since at least the 1930s, but there are important differences. First, only the EDA channel need be used with the GKT. Examiners may record the other two traditional polygraph channels to comply with licensing, regulations or APA requirements when conducting a GKT, but they need not be interpreted. Second, the test items are in random order, except that the first item is always a buffer item. Third, the GKT is scored objectively, using a method described later in this paper.
Before taking up a discussion of the GKT procedure, let us first address the natural question as to the reason that a trained field polygraph examiner, highly competent in conventional techniques, would consider using the GKT. One reason is that there is more general agreement in the scientific community regarding the defensibility of the theory underlying the GKT. While scientific agreement doesn’t necessarily solve crimes, it does have bearing on whether a testing method meets evidentiary standards for courtroom admissibility. A polygraph examiner wishing to use the results of a polygraph examination as evidence in court would find more scientific support with the GKT than any other polygraph technique. This is not to imply that the other methods are less valid, only that many scientists have a preference, and that preference is the GKT.
The GKT is very simple to administer. Before the examination, the examiner constructs a series of test question lists where the items are taken from the case facts. The best way to explain the test construction process is to use an example. Let’s say that the St. John Pharmacy was robbed by a man wearing a mask over his face and holding a screwdriver. He took $1880 during the holdup. In addition, he stabbed the owner in the hand. The robber then ran down the street, and escaped by jumping on a bus that had pulled up to the corner.
We can assume that many of the details of that robbery would be remembered by the thief: the name of the business, the weapon, the disguise, the amount of money stolen, where the owner was injured, and the form of transportation used in the escape. Similarly, an innocent person would probably not know them unless there had been a leak of information to the public. Assuming that the police had properly withheld these details, a GKT could be administered using the crime information on any suspect brought in for testing.
Each case fact would have a separate test. Using the present example, the type of weapon would make a good GKT. The following items might be used: pistol, rifle, machete, knife, broken bottle, and of course, screwdriver. The screwdriver would be called the critical item, or sometimes the “key”. All items in the list must be equally plausible: the innocent must not be able to guess which is the critical item in the list because of some characteristic of the item. In addition, there should not be any implausible items on the list, i.e., a cannon or a noose. It is also important that the critical item be something that the guilty person is very likely to remember. The robber may not recall the color of the storeowner’s shoes or hair because they were less important to the crime. One should also be mindful that crimes are often committed by intoxicated criminals, whose memories are compromised by drugs or alcohol. Care must be taken in selecting the critical items for the GKT that are very likely to be recognized by the guilty examinee.
Each GKT list should have 5 or 6 items, including the critical item, and there should be 3 or more lists, each with a different critical item. The placement of the critical item among the other items in a list should be done at random, using a random numbers table or some other means. However, the critical item must not be in the first position. That position is always reserved for a neutral item. The reason for having 3 or more lists of items will be discussed later in this paper.
Once the lists are made, the examinee can be tested. The sensors are placed in the usual manner, but as stated previously, only the EDA channel will be scored. The examiner than begins the test. A question series using the type of weapon could be presented like this:
“If you are the robber of that business, you know what type of weapon was used. Just repeat the weapons I say after me. Was it a: pistol, rifle, machete, knife, screwdriver, broken bottle?”
The examinee is instructed to repeat the items to ensure he is paying attention. Since only the EDA is used, question spacing can be closer than the 25 seconds recommended for CQT testing. The GKT is given only once per list, and then another list is used to conduct another GKT.
After 3 or more GKTs are given, the examiner can begin to score them. Scoring is very simple. The first item on the list is never scored, only the remaining items. Using only EDR amplitude, if the strongest EDR occurs on the critical item, it receives a score of 2. If the second highest is the critical item, the score is 1. No score is given if the EDR to the critical item is ranked third or lower. Therefore, the highest score possible for 3 GKTs is 6, and the lowest is 0. Decision rules vary, but Lykken (1959) reported success using a cutting score that was equal to the number of GKTs administered. In the present example, a score of greater than 3 would indicate the examinee had knowledge of details of the crime not expected of an innocent person.
The real beauty of the GKT, in addition to its simplicity, is that the test permits the user to calculate the precise probability of a false positive error. The capability to calculate the error of a test is a factor many jurisdictions find important in forensic testing. One method of error estimate for a version of the CQT has shown promise (Dutton, 2000; Krapohl & McManus, 1999), however at this writing the GKT has the larger body of supporting data. Probability tables using a similar ranking protocol for GKT scores are found in Dr. Lykken’s recent book (1998, pp 290, 292).
The reason that it becomes important to have 3 or more GKTs has to do with probabilities. The likelihood that an innocent person would react largest to the critical item is a function of the number of items. If there were 5 items on the test, excluding the first item, the probability that an innocent person would react greatest to the critical item is 20% (1 in 5). The likelihood of that same innocent person reacting to the critical item on the second GKT is also 20%, and so on through the number of GKTs. If the examinee reacted on the first and second GKT, the probability that this occurred solely by chance would be 4% (20% of 20%). Reacting also on the third GKT would be even more unlikely for the innocent person: 0.8% (20% times 20% times 20%). In other words, there is about one chance in 100 of an innocent person, unaware of the true crime details, would react most strongly to the critical items on all 3 GKTs. Reacting to the critical item on all subsequent GKTs further reduces the chance probabilities by 20% each.
One of the limiting factors for the GKT is the availability of critical test items. The success of the GKT depends on investigators having access to crime details that have not been released to the public or the examinee. In contrast, the CQT can be used in almost all circumstances, since the technique does not depend on the availability of restricted information. Podlesny (1993) conducted a file review of polygraph cases conducted by the Federal Bureau of Investigation (FBI) in order to determine to what extent the GKT could be used by the FBI in its crime investigation mission. His data indicated that only a small portion of the FBI polygraph caseload was amenable to the GKT methodology, a mere 13.1%. This percentage may or may not apply to other settings, where crime details are handled differently by the police and the news media. Podlesny suggests that the GKT might be introduced as an adjunct to the CQT testing, in order to reduce errors, but that it does not offer a strong alternative to the CQT in most field applications.
In summary, the GKT is an elegant and powerful testing technique, with many important advantages. Its only significant shortcoming is that it cannot be used in all cases, but only those where there are a sufficient number of critical items. Examiners are encouraged to try the GKT, to become familiar with it, so that they have yet another tool at their disposal.
References
Crane, H.W. (1915). A study in association reaction with an attempted application of results in determining the presence of guilty knowledge. Psychological Monographs, 18(4), 1-61.
Dutton, D.W. (2000). Guide for performing the Objective Scoring System. Polygraph, 29(2), 177-184.
Krapohl, D.J. & McManus, B. (1999). An objective method for manually scoring polygraph data. Polygraph, 28(3), 209-222.
Lykken, D.T. (1959). The GSR in the detection of guilt. Journal of Applied Psychology, 43, 385-388.
Lykken, D.T. (1998). A tremor in the blood: Uses and abuses of the lie detector. Plenum Trade: New York.
Podlesny, J.A. (1993). Is the guilty knowledge polygraph technique applicable in criminal investigations?: A review of FBI case records. Crime Laboratory Digest, 20(3), 57-61.
Acknowledgments
We are grateful for the editorial review and technical information provided by Dr. Gordon Barland. Reprinted with permission. This article originally appeared in Spanish on the Web site of the Latin American Polygraph Association (www.lafayetteinstrument.com/alp/entrada/alp.htm). Donald J. Krapohl is a member of the South Carolina Polygraph Association (SCAPE), American Polygraph Association (APA), American Society for Tests and Materials (ASTM), and the Society for Psychophysiological Research (SPR). Yazmín Vélez is a polygraph examiner, Director of International Sales with the Lafayette Instrument Company and a member of the APA and the Latinamerican Polygraph Association. The first author can be contacted at dkrapohl@aol.com, and the second at yvelez@licmef.com.